Skip to main content

Full text of "Further studies of the bleeding of Portland cement paste."

See other formats


7 194F 



Research Laboratory 

of the 



Portland Cement Association 



UJ 






Bulletin 1 









P 
m 






Further Studies of 











■ 



Urn 



i 



The Bleeding of Portland Cement Paste 




aa 



BY 



HAROLD H. STEINOUR 



December. I9-J 



Chicago 



Research Laboratory 

of the 



Portland Cement Association 



Bulletin 4 



Further Studies of 
The Bleeding of Portland Cement Paste 



BY 



HAROLD H. STEINOUR 



December, 1945 



Chicago 



Foreword 



By T. C. Powers 



In July 1939 an extensive paper on bleeding was 
published — Bulletin 2 of the Research Laboratory. 
Although the work behind this paper improved our under- 
standing of the phenomenon, showing bleeding to be a 
special case of sedimentation, it left open several 
questions of practical and fundamental importance. 
Particularly, the theoretically derived equation, which 
attempted to show the relationship between the various 
factors controlling the initial rate of settlement, 
contained an empirical factor that could not be wholly 
interpreted. This empirical constant seemed to embody 
the effects of those characteristics other than specif- 
ic surface which distinguish the behavior of one mate- 
rial from that of another, but there was also the pos- 
sibility that this constant served to compensate for 
certain fundamental misconceptions embodied in the 
equation itself. Consequently, the subject could not 
be considered closed until something more was known 
about the significance of this constant. 



When Dr. H. H. Steinour joined the laboratory 
staff in January 1940, he undertook the project of 
clearing up this question. By much careful and sys- 
tematic work he has succeeded in increasing our under- 
standing of the bleeding of cement pastes to the point 
where, so far as additional research is concerned, the 
subject can be closed until new reasons for continuing 
the study arise. 

In this paper Dr. Steinour 1 s extensive studies of 
the physical factors that influence bleeding and settle- 
ment in general are presented in summary only. Three 
complete papers on the fundamentals were published in 
Industrial and Engineering Chemistry (July, September, 
October, 1944) • These papers set forth laws that should 



3 -in- 



apply to chemical engineering both In and outside the 
cement industry, particularly to the process of filter- 
ing or to thickening by sedimentation* 



The present paper sustains the earlier conclusion 
that specific surface is the most important factor con- 
trolling the bleeding characteristics of a portland 
cement. It goes beyond Bulletin 2 in showing that for 
cements of given specific surface the bleeding charac- 
teristics are controlled by the shape of the particles, 
by their state of flocculation, and by chemical reac- ] 
tions that occur during the first few moments after the 
cement and water make contact. It points out also 
that in abnormal cases bleeding is controlled by the 
same factors found by William Lerch to control premature 
stiffening. All these factors are subject to control, 
although some of them, especially the shape of particle 
and degree of flocculation, may not be controllable on 
a commercial scale. 



One point made clear by the extended studies is 
that there is no hope of finding a universal simple 
formula for predicting bleeding characteristics of a 
cement from its composition. However, the work pro- 
vides an understanding of the nature of the variables 
involved and their effects, and it shows how to test 
any given material. 



On the basis of these studies the writer is now 
inclined toward the view that if any steps at all are 
required for the control of the bleeding characteristics 
of a given cement they should be confined to a regulation 
of the specific surface of the cement or perhaps to the 
addition of air-entraining or gas-forming agents; how- 
ever, neither a change in specific surface nor any addi- 
tion should be made without a careful study of the ef- 
fects on all the important properties of concrete. With 
respect to other factors that influence the bleeding 
characteristics, it now seems that differences in bleed- 
ing characteristics among cements of similar specific 
surface are not fundamentally of great importance; they 



- iv - 



are decidedly secondary to other properties that the 
manufacturer must control. 



In the Introduction to Bulletin 2 (p. 2) the view 
was expressed that there was an important connection 
between the bleeding characteristics of the original mix 
and the durability of concrete when exposed to weather- 
ing agencies, and that a reduction in bleeding was in 
the direction of improvement. This view seemed to be 
shared at the time by various investigators of concrete, 
but nothing has developed since to sustain it. On the 
contrary, some data show that among concretes of rela- 
tively low durability those which showed the greatest 
amount of bleeding gave the best resistance to deterio- 
ration by freezing and thawing. 



On the theoretical side we have gained a new under- 
standing of the nature of pastes, tfe now know that a 
Portland cement paste or any other paste of highly con- 
centrated, flocculated particles has a continuous struc- 
ture which, in the period before hardening, shows some 
elasticity and much plasticity*. At the same time it 
seems that the articulation of this structure is effected 
by forces acting across thin layers of water that sepa- 
rate points of near contact between individual particles. 
The establishment of these two facts makes it possible 
to understand and explain several aspects of the proper- 
ties and behavior of fresh cement pastes, especially 
the plasticity. It also reveals certain current fal- 
lacies, particularly the deductions based on a popular 
but erroneous conception of the flocculated state of 
cement particles. 



Dr. Steinour's work upholds the basic conceptions 
that were set forth in Bulletin 2 but at the same time 
shows the need for modifying some of them. For example, 
the basic equation for the initial rate of bleeding was 
found to be not wholly consistent with the concept under- 
lying its derivation. Also, the theoretical equation 
for bleeding capacity was supplanted by an empirical 



- v - 



equation that covers a much wider range in water con- 
tent than did the theoretical one« Moreover, the theory 
tentatively put forth in Bulletin 2 concerning the role 
played by liquid-adsorption was found not to be tenable; 
instead, the extent of initial chemical reaction was 
found to be a significant factor. 






- vi - 



FURTHER STUDIES OF 
THE BLEEDING OF PORTLAND CMENT PASTE 



by Harold H. Steinour 



Table of Contents Page 



Abstract 



1 



Introduction 



1 



1- The Dormant Period 3 

2. The Initial Bleeding Rate and Its Theoretical 

Significance 7 



The Rate Equation of Bulletin 2 7 

Basic Studies on Rate of Sedimentation . .. 9 

Comparison of Rate Equations 13 

Probable Causes of Differences in the 

Bleeding Rates of Different Cements • • • • 13 



3» The Bleeding Capacity Considered in Relation to 

Water Content of Paste 16 



Theoretical Aspects • 16 

Snpirical Relationships 19 

Implications of Empirical Relationships . • 21 



/+• Relationships between Cement Fineness and the 

Bleeding Characteristics 2U 



Bleeding Rate 24 

Bleeding Capacity 26 

5« Relationships between Bleeding Rates and Bleed- 
ing Capacities 28 

The Experimental Evidence 28 

Theoretical Considerations 31 






- vii 



6. Duration of Bleeding 



Page 

33 



7. The Dominant Chemical Influences in Cements of 

Normal Production •••••••••••• 35 






Statistical Measures 35 

Heat Liberat ion 39 

Readily Leached S0 3 40 

Soluble Alkalies 42 

Apparent Increase of the Density of 

Cement in Contact with Water 42 

Other Properties 43 

45 



Summary 

8, Miscellaneous Influences Affecting Bleeding Rate 



and Bleeding Capacity U5 



Rate of Clinker Cooling 45 

Calcium Sulfate Ret arder U& 

Aerat ion of Cement U9 

Salts U9 

Alkali Hydroxides 51 

Temperature During Test 54 

Vinsol Resin 56 

9. Summary and List of Factors That Affect the 

Bleeding of Cement Paste 57 



Nomenclature 61 



References 



63 



Appendix A: Chemical Constitution 65 

Appendix B: Bleeding Data for Pastes Made with 

Cements of the SBR Group 69 



- viii - 



Page 

Appendix C: Additional Study of Equations 2 and 5* 73 

Appendix D: Mathematical Developments with Respect 

to Bleeding Capacities 77 

■ 

Derivation of Equation 6 of Section 2 ... 77 
The Assumed Linear Relationship between 

VzSF and w,for Coarse Cements 77 

Appendix E: The Measurement of Bleeding of Pastes. 81 

Suitable Equipment 82 

Preparation of Paste &U 

Measurement of Bleeding 85 

The Bleeding Data o 86 

Cement Density •• 87 






- ix - 



FURTHER STUDIES OF 
THE BLEEDING OF PORTLAND CEMENT PASTE 



by Harold H. Steinour 



Abstract 



Results of studies on the bleeding of cement 
pastes are reported for the period since publi- 
cation of Bulletin 2 of the P.C.A, Research Labo- 
ratory. Basic work on the general problem of 
the sedimentation of concentrated suspensions is 
published in more detail elsewhere but is sum- 
marized here. On the basis of accumulated evi- 
dence, empirical relationships between bleeding 
capacity and water content of paste are presented 
which have wider ranges of applicability than 
the relationship previously assumed. Relation- 
ships between bleeding rates and bleeding capa- 
cities, and such aspects as the effect of cement 
fineness and the duration of bleeding are also 
discussed. Bleeding data for a number of cements 
are compared with other properties of the ce- 
ments, and correlation between the rate and 
amount of bleeding and the early chemical reac- 
tivity of the cement is shown. Miscellaneous 
test data which indicate that bleeding rate and 
bleeding capacity are subject to many influences 
are presented, A method of conducting the bleed- 
ing test is outlined as an appendix. The report 
deals only with the bleeding of pastes, not 
mortars or concretes. 



Introduction 



This report is supplementary to Bulletin 2 issued 
by the Research Laboratory of the Portland Cement Asso- 
ciation in 1939 under the title, "The Bleeding of Port- 
land Cement Paste, Mortar and Concrete," by T. C. Powers. 



-1- 



interim^ 
tinued and numerous tests have been made both on ce 



ment pastes and on concretes. In addition, tests have 
been made on suspensions of other solids, in further 
investigation of theoretical aspects.* The present 
report reviews the general implications of this recent 
work with respect to the bleeding characteristics of 
cement pastes. 

The bleeding of a cement paste may be described as 
the development of a layer of water at the top of the 
paste as a result of the settlement of the cement par- 
ticles. The type of testing employed in the work covered 
in Bulletin 2, and in that which has been done since, 
consists in obtaining a time record of the level of the 
paste at the center of a wide vessel. The observations 
are made with a micrometer microscope by sighting on a 
"float" which remains at the paste-water boundary. Prac- 
tical details supplementing those given in Bulletin 2 
will be found in Appendix E. 

Most of the cements for which data are reported 
have been given either the group identification SBR or 
LTS. The cements in both groups were produced fr 
commercial clinker and nearly all of them were ground 
in commercial mills. Chemical compositions are given 
in Appendix A. The letters relate solely to the re- 
search purposes for which the cements were originally 
obtained. They stand for Special Basic Research and 
Long-Time Study. 



Most of the data given in this report for cements 
of the SBR group were obtained under a program of basic 
research supervised by T. C. Powers. Early studies by 
William Lerch supplied the first evidence for some of 
the chemical effects. Data obtained from W. C. Hansen 



*A part of this work is published in Industrial and 
Engineering Chemistry under the title "Rate of Sedi- 
mentation," * See references (14), (15) and (16). The 

papers have been reprinted as Bulletin 3 of the Research 
Laboratory of the Portland Cement Association. 



-2- 



on cements of the LTS group were utilized in the general 
comparison of bleeding characteristics and other prop- 
erties of cements. Most of the bleeding tests were 
made by Robert H. Borkenhagen, Lynn A. Brauer, Richard 
G. Brusch, Frank Rosal, Herbert W. Schultz, and Edwin 



M. Wiler. 



1. The Dormant Period 



It was found experimentally and was reported in 
Bulletin 2 that for a considerable time after a neat 
cement paste is mixed and placed in a deep, wide vessel 
the rate of subsidence of the paste at the center of the 
vfiflsftl remains strictlv constant (as nearly as can be 



1 second). 



.001 mm 
heights 



that were conmonly used, the constant-rate periods some- 
times lasted 30 minutes or more. A special test indi- 
cated that under suitable conditions the constant rate 
could be maintained for 75 minutes. It was observed, 
also, that the bleeding capacities, or total settlements 
expressed as proportions of the initial heights of paste, 
were independent of the initial heights for total bleed- 
ing times up to 75 minutes. Consequently, it was con- 
cluded that for 75 minutes most cements show little 
chemical reactivity and may be considered to be in a 
dormant state, at least as far as effects on bleeding 
are concerned. (See Bulletin 2, pp. 7U> &U») 

As a part of the subsequent work, further tests 
were made to determine the maximum periods during which 
various cements would maintain constant bleeding rates. 
The final estimates were obtained from bleeding tests 
made in cylinders 9 cm in diameter, filled with cement 
paste to a depth of 28 cm. Five cements of average 
fineness, representing different clinkers, were tested 
in this way. For two of the cements, wall-resistance 
appeared to be negligible and the test consisted simply 
in determining for how long the initial constant-rate 
period was actually maintained. In the other cases a 
change from the initial to a slightly lower rate occurred 
after 20 to 40 minutes. This second rate remained 



-3- 



practically constant for a considerable time and than 
gradually diminished. The change from the one constant 

slichtlv lower one was thought to be 



rate to the other, 



>istance since, in preliminary tee 
narrower cylinder, more such chang 



were often clearly evident. It was concluded that the 
successive constant rate periods should be added to- 
gether to estimate the time for which a single constant 
rate might be maintained in the absence of wall effects; 
for the same cements this gave about the same results 
irrespective of whether the wide or narrow cylinder was 
used. Fortunately, the two extremes of the range in 
dormant periods were found in those teste in which no 
allowance for wall resistance was needed. The results 
for the tests in the 9 cm-dia. cylinder are" shown in 

Table 1. 



In the preliminary work with the narrower cylinder, 
cements of widely varied fineness (a series for each 
clinker) had been tested and had shown no large or con- 
sistent effect of fineness on the duration of the dormant 
period. Apparently, most cement pastes prepared as in 
these tests can maintain, under suitable conditions, 
remarkably constant bleeding rates for periods of 2/3 
to 2 hours. 



A somewhat different approach to the study of the 
dormant period was adopted in other tests, in which 
pastes of a given concentration were allowed to stand 
after mixing i for intervals of successively increased 
length, and were then briefly remixed and tested for 
bleeding at paste heights of only a few cm, as in the 
usual bleeding tests. Data on the most extensive test 
of this kind are shown in Table 2. The cement used to 
obtain the data of Table 2 had shown, in tests in the 
tall 9 cm-dia. cylinder, a dormant period of about 60 
minutes. This agrees in part, at least, with the data 
of the table in that the bleeding rate apparently first 
began to be reduced significantly in the test that was 
begun after the 60-minute rest period. The table is 
believed to show in a rather striking way the constancy 



-4- 



Table 1 



Estimated Dormant Periods and Other Data 
for Testa on Pastes 28 cm High f in Cylinders 

9 cm in Diameter, at 23»5 a C 



Times are figured from the end of a 7 -minute mixing schedule: 

2 1 mix - V wait - 2 f mix. 
The proportions of water in the pastes were 0*575 to 0*600 

by absolute volume* 
The bleeding rates in millionths of a cm per sec were 125 to 

150 # No channels or boils developed,* 



Cement 
Lot 


Clinker 
Ref. 
No. 


Total** 
Spec. Surface 
( Wagner Turb . } 
cm £ /gjn 


No. of 
Const. 
Rate 
Periods 


Apparent 
Dormant 
Period, 
min* 


Total 
Bleeding 
Time, 
min. 


SHR-15365 


1 


I665 


2 


60 


116 


SBR-15A96 


2 


1740 


1 


40 


81 


SHR-15621 


3 


1820 


2 


60 


131 


SHR-15668 


k 


1830 


1 


, U5 


205 


SBR-15697 


5 


1705 


2 


90 


161 



*For the purpose of minimizing wall resistance, faster bleeding pastes 
obtained by increasing the water contents would, have teen preferable. 
It was found, however, that, because of the heights used,* pastes 
having greater water contents were susceptible to channeling* 

**This is the value obtained by the A.S.T^M. method, plus an estimate for 
the surface area of the residue on the 325-mesh sieve* 



-5- 



Table 2 
Bleeding Tests on Remixed Paatea 

Cement ITS -21 

Water content of paste: 0.601 by absolute Tolume 

Initial mixing schedule: 2* mix - 3 f wait - 2 f mix 

Final mixing: \ min. 

Height of paste: Jb mm 

Temperature: 23.5°C 

Results for remixed pastes are averages of 2 or 3 tests 



Rest 


Bleeding Rate, 


Bleeding 


Duration of 




Time between 


Period, 


Q, 


cm per 
x 10 6 


sec, 


Capacity, 


Const. Rate 


Duration of 


Initial Mix 


min. 






AH' 


Period, min. 


Bleeding; 


and End 












(Approximate) 


w 

min* 


of Bleeding 







194 




0.122 


20 


55 


55 


15 




189 




0.113 


18 


55 


70 


30 




196 




0.125 


15 


57 


87 


45 




19? 




0.106 


17 


52 


97 


60 




185 




0.103 


18 


48 


108 


90 




172 




0.090 


20 


47 


137 


120 




167 




0.075 


12 


45 


165 



Symbols used conform to those of Bulletin 2. 
list at the end of this report . 



See also the nomenclature 



-6- 



with wnich bleeding properties can be maintained for a 
considerable time. 

Though the period during which a constant bleeding 
rate can be maintained has been called the dormant pe- 
riod, it appears from other data that chemical reactions 
are not strictly in abeyance during this time. A dis- 
cussion of how these different aspects may be reconciled 
is given at the close of the next section. 

2. The Initial Bleeding Rate 
and Its Theoretical Significance 

T he Rate Equation of Bulletin 2 

In Bulletin 2 much of the theoretical analysis re- 
lates to the initial period of constant bleeding rate. 
It was assumed that during this period the cement par- 
ticles in the top layer of the paste maintain the ar- 
rangement and spacing that results from the mixing, and 
that onposition to the settlement of these particles is 
caused only by the viscous resistance of the free water 
in the spaces between particles. By analogy to the work 
of Fair and Hatch (7) and of Kozeny (9) and Carman U,5) 
on viscous flow through granular beds, it was further 
assumed that the flow of the water between the settling 
particles conforms to Poiseuille's law for capillary 
flow, as adapted to non-circular channels by use of the 
hydraulic radius.* By employing these assumptions, using 
a proportionality constant determined empirically by 
Carman, an equation for the initial constant rate of 
bleeding was derived. When all the water in the paste 
is assumed to take part in the flow between the par- 
ticles, the equation takes the following form: 

0.2 g( d c- d fj_ w3 m** 



'he hydraulic radius may oe aenneu us ouo xc^xw ~* 
volume of moving fluid to the area of the surf 
wetted by that flujld t 

'his is the expanded form of Equation H of Bullet 
when w^ = 0'« 



'7. 

i 



where : 



Q ■ the rate of bleeding, cm of subsidence per 
second 



* the gravitational acceleration, cm/sec 

d c = density of cement, gm/cm 

df - density of the water, gm/cm 

n - viscosity of the water, poises 

2/3 

CT = specific surface of cement, cm /cm 

w = volume of water per unit volume of mix 

c = volume of cement per unit volume of mix 
(c + w = 1) 



Bui 



let in 2, Powers found 
as follows: 



n _ 0.2 g( d c- d f) ( w-w< ) 3 (2) 

^ ~ <? W 2 * ~ ♦ . C 

where the specific surface in sq cm per cu cm is here 
written as o~ w to indicate that it was determined by the 
Wagner turbidimeter method, and 

Wj^ ■ a constant for tests on a given cement at a 

given temperature, which, since it is sub- 
tracted from w, was assumed to represent a 
quantity of immobile water per unit volume 
of paste. 

The empirical constant, w^, was found to have dif- 
ferent values for different cements, but it was always 
of such size that had it not been introduced the esti- 
mated bleeding rates would have been much too high, 
often 6 or 7 times the actual rates. Since the physical 
significance of the term remained largely a matter of 
speculation, the need for w± constituted a weak link in 



""Same as Equation H of Bulletin 2. 



-8- 



the evidence supporting the theory of bleeding that had 
been developed. — 

Water of hydration was an insufficient explanation 
of w it for a v^ term was needed also in representing data 
obtained for pastes of cement and kerosene, and for other 
inert systems. The possibility of stagnant liquid was 
considered but seemed to be ruled out by the work of 
Carman (4,5), who found no need of a w^ term in analyzing 
the flow of liquid through beds of particles (of low 
specific surface). The explanation of w i that was ten- 
tatively offered in Bulletin 2 was developed in terms of 
adsorption of liquid by the solid. Though this explana- 
tion agreed with views of von Buzagh (3) (who was quo- 
ted), there are other investigators who believe that ad- 
sorption layers of the thicknesses necessary to account 
for the w i values are not possible (1,2). It is also 
to be noted that the magnitude of w A is independent of 
the quantity of cement in unit volume of paste. This 
fact is somewhat difficult to reconcile with the inter- 



imply 



unit volume 



Further work on the nature of w i seemed advisable 
both for its bearing on the theory and for the insight 
which it might give with respect to the different bleed- 
ing tendencies shown by different cements. The three 
papers that have since been written on "Rate of Sedimen- 
tation" (14,15,16) show how the problem was attacked 
through work on simpler systems. Here only a brief 
survey of some of that work will be given. 

Bas i c Studies on Rate of Sedimentation 

Work with uniform spheres in the disperse, or non- 
flocculated, state showed that for concentrated suspen- 
sions of the spheres the following rate equation was 



satisfactory: 



_ 0.22,6 g ( d c- d f) _w 3 



(3) 



-9- 



where the symbols have the same significance as given 
previously. except that "solid" and "fluid" should be 
read in place of "cement" and "water". It will be noted 
that this equation conforms to Equation 1 except for the 
value, 0.246, of the empirical proportionality constant. 
It therefore indicates substantial agreement of the basic 
theory of Bulletin 2 with data for uniform, non-floccu- 
lated spheres. 

However, to obtain agreement with data for a set of 
tests on chemically inert, uniform-size, angular par- 
ticles in non-flocculated suspension the equation had to 
be changed to 



0.36 g ftc-df) (w-0.169) 



3 



That is, not only was a term needed like the w^ of Equa- 
tion 2 but the constant 0.2^6 had to be replaced by 0. 36. 



By assuming immobile liquid, it was possible tc 
count for both changes. Any immobile liquid in a 
flocculated system, such as was under study, woul 
located at the individual particles of solid, and 
quantity per unit volume of the suspension would 
sumably be proportional to the solid content, c. 



Cor- 



rection for such immobil 
transforms Eauation ? to 



^1 Cf 

1-Wi 



q = . Q'ffi , g( d c- d f) fesaJ (c) 

if use is made of the fact that c « 1 - w. When w^ 
= 0.168, the factor 0*246 amounts to O.356. Hence, 

(l-WjJ 2 

Equation 5 as applied to the systems under study shows 
excellent agreement with Equation 4. The assumption of 



*Th 



t m m i 



and 



Lively, the mobile and imm 

* 

See also Appendix C. 



* II 



-10 



immobile liquid proportional to c is thus wholly adequate 
to explain the results, and because of the simple nature 
of a non-flocculated suspension of uniform-size particles 
the possibility that there may be other equally satis- 
factory explantions seems small. 

Experiments with different-size particles, and study 
of published work, led to the conclusion that in non- 
flocculated suspensions of inert particles the magnitude 

is independent of particle-size and dependent 

solely on particle shape* 

The results of the work on non-flocculated sus- 
pensions may be summarized as follows: In a non-floc- 
«„t«4-^^ anc-rxarxsHriTi nf nTvrrrrnrimatelv uniform-size parti- 




theory 



agree 



with the equation except for the value of the proportion- 
ality constant which is, after all, empirical. When the 
particles are angular, they apparently carry with them a 
quantity of liquid which, per unit volume of the suspen- 
sion, is proportional to the volume of solid and indepen- 
dent of particle-size. The large quantity of the liquid 
involved, its invariance with respect to the degree of 
subdivision of the solid, and its dependence on particle 
shape are considered to be strong indications that the 
liquid is not bound to the particles but simply remains 
stagnant at angularities in their contours. That is, the 
sharp angularities of the particle-surfaces apparently 
give rise to dead spaces where the liquid remains rela- 
tively immobile. 

In Bulletin 2, stagnant liquid was not assumed be- 
cause Carman had found no indication of it in studies on 
flow through uniform granular beds. The experimental 
results discussed here appear to demand this assumption, 
but even if the original explanation of w t in terms, of 
adsorbed liquid could still be made, it would no longer 
avoid disagreement with results obtained by Carman for 
granular beds, for Carman has now reported work with 



11- 



fine powders (6) and has in general found no need to 
assume an appreciable adsorption. Apparently, flow of 
liquid past suspended particles is not as strictly anal- 
ogous to the flow through the compact beds investigated 
by Carman as was at first thought. This is indicated 
also by the change in the proportionality constant of the 
rate equation. 

In the further course of the investigation of rate 
of sedimentation, tests were made on flocculated suspen- 
sions of the same angular particles that had been tested 
when not flocculated. Equation $ was still found to 
apply, to a good degree of approximation, but the value 
of Wj^ was greater than before. That is, the suspensions 
settled more slowly when flocculated. Two possible ex- 
planations are seen for the increase in w^: (1) The quan- 
tity of stagnant liquid may be increased by reason of 
interparticle contacts caused by the flocculation. (2) 
There may be small isolated pockets of liquid distributed 
through the flocculated mass. Such pockets might con- 
tribute to the value of w^ in approximately the same 
way as stagnant liquid. That is, liquid removed from 
the floc-space in this way would lower the proportion 
of water in the latter, and the correction term would 
be of the same form as for water that is stagnant at 
the individual particles.* 

Many different powders, each containing a wide range 
of particle-sizes, were tested as flocculated suspen- 
sions. Most of them showed fair agreement with Equation 
5, thus demonstrating that agreement with the equation 



un 



particle. 



To test directly the assumption tnat only the vis- 
cous resistance of the water opposes the fall of the 
flocculated particles during the constant-rate period, a 
number of tests of hydrostatic pressure were made, prin- 
cipally on cement pastes. The results agreed fully with 
the assumption. 



*3ee brief explanation in the second article on "Rate of 
Sedimentation" (15). 



-12- 



Comparison of Rate Equations 

Attention needs now to be directed to the fact that 
Equation 5, which was developed from this basic study, is 
not strictly the same as Powers 1 Equation 14, which is 
represented by Equation 2 of this report. This might 
lead to the supposition that cement pastes do not behave 
like these other flocculated suspensions. However, it 
has been found (16) that for cement pastes Equations 2 
and 5 give results that agree within the limit of error 
of the data. Evidently, therefore, the cements act like 
the majority of the other powders, and the ideas that 
have been developed regarding the significance of w± are 
as applicable to the cements as to these other powders. 

As a result of the studies just discussed, Equation 
2 as applied to cement pastes is now regarded simply as 
an approximate version of Equation 5, in which a constant 
times a w is used in place of the more accurate O, and 

1 is represented by an average value appropri- 

(1-Wi)" 

ate to the particular limited range of Wi values found 
for cements. Since the approximations seem to be no 
greater than is justifiable in practical work, and since 
they simplify both the experimental work and the compu- 
tations and give results directly comparable with those 
in Bulletin 2, Equation 2 was used for the determination 
of all values of w^ for cements that are cited later in 
this report. Indeed, accurate values of a necessary for 
a strict application of Equation 5 were not available 
for these cements. 

Probable Csuses of Differences in the Bleeding Rates of 
Different Cements 

The contributions to the v^ values that have been 
discussed thus far may be considered to be wholly of 
physical origin, for the studies were made on chemically 
inert systems. In such systems, as has been seen, the 
magnitude of w. is apparently determined by the particle 
shape and by the state of flocculation. In cement paste, 



-13- 



however, it seems doubtful whether the differences in 
particle-shape are sufficient to contribute importantly 
to the differences in the w^ values. The cement pastes 
may differ somewhat in the magnitudes of their floccu- 
lating forces, but, again, it is doubtful if the w^ 
values are much affected. The cement pastes are all 
flocculant, and when channeling is avoided, the value of 
w^ is not sensitive to moderate differences in the 
strength of flocculation. This is indicated by the 
studies on silica powder that are reported in Bulletin- 2 
(p. 72). Values of w.^ do vary markedly with one physical 
property of cements, namely, the fineness. However, 
tests on flocculated suspensions of inert powders (16) 
showed little effect of fineness on w^ when the powders 
were composed of many different sizes of particles, as 
are portland cements; therefore, the differences in the 
Wi values found for cements are probably mainly a result 
of chemical reaction. 

Indeed, if significant amounts of reaction products 
form as coatings on the cement particles during the mix- 
ing period , the different reactivities of different 
cements provide a ready explanation for all the dif- 
ferences in Wi values. Such coatings would increase the 
immobile phase at the expense of the mobile one, and this 
would increase w i# Moreover, under the assumption that 
the thickness of the coating would not vary much with the 
fineness of the cement, the volume of the immobile phase 
would increase with the fineness, thus accounting for the 
increase in w^ with fineness. 

Arguments that were advanced in Bulletin 2 (p. 56) 
against the assumption that there is appreciable forma- 
tion of reaction products before the end of the dormant 
period now appear less forceful as a result of the subse- 
quent studies. The proportionality constant of the rate 
equation is no longer considered to be maintained so pre- 
cisely at just the same value for different cements as 
was formerly thought. Neither Equation 2 nor Equation 5 
is so precise a representation of the data for cements 



-H- 



as to rule out the possibility that an appreciable for- 
mation of reaction products occurs during the mixing 
period. Indeed, the necessity for a long and vigorous 
mixing (see Bulletin 2, p. 135) can be regarded as sup- 
port for the assumption that a significant amount of 
initial reaction does take place. The initial reaction 
need not cease entirely after the mixing period, for a 
reaction layer initially formed loosely over the parti- 
cles could fill in without altering the bleeding rate. 
Also, any reaction product that was formed in, or swept 
into, the small, stagnant spaces that are apparently to 
be found at angularities in the particles would have no 
effect on the rate. Heat-data, illustrated by Fig. 1, 
are consistent with the views on chemical reaction that 
have just been stated. Microscopic observations also 
give some indication that a film of hydration product 
does form (10). Such coatings need not be more than 
a few tenths of a micron in thickness in order to ac court; 
for the differences in the v ± values. The dependence of 
bleeding characteristics upon specific chemical factors 
will be considered in later sections of this report. 

Because of the changed conception of the signifi- 
cance of w t , the original definition of this term needs 
modification. Even when stagnant water and water of hy- 
dration are wholly responsible for w^, the quantity of 
such water per unit volume of paste is not equal to vi ± 

but to Wi c. Because of this, and also because any 

l-w i 
water in "pockets" contributes to *i in a different way, 
the term does not represent one simple physical entity. 
It can be defined as a correction for liquid that is not 
involved in the flow, but it cannot be considered to rep- 
resent, directly, the quantity of such liquid. 

Fig. 2 shows the ranges in bleeding rates that are 
likely to be found at various water-cement ratios when 
cements are of ordinary, but somewhat variable, fineness 
and show the usual variations in w^. 



-15- 



i- 

o a 35 




4 5 

Time -hours 



Fig. I -Rate of Heat Liberation of Cement Paste versus Time 




1.6 1.8 2,0 

The Water-Cement Ratio by Absolute 



Vo 



ume, » 



Fig. 2- Ranges in Bleeding Rates 

Most cements with specific surfaces of 1850 ± 100 
sq. cm. per gm. will give pastes whose rates of 
bleeding fall within the band shown. The different 
rates found for different cements are attributable 
to differences in chemical constitution and in 
treatment as well as to differences in fineness. 



3. The Bleeding Capacity Considered in Relate 

to Water Content of Paste 



Theoretical Aspects 

Bleeding capacity, or the total settlement per unit 
of original paste height, increases with the initial water 
content of the paste. An equation for this relationship 
was developed in Bulletin 2 (p. 65) on a simple theoreti- 
cal basis, and was shown to agree with experimental data 
to a considerable extent. However, because of new exper- 
imental data there is need to reconsider both the theory 
and the assumed quantitative relationship. 

In the original theory it was assume.d that if a 
cement paste could be consolidated by causing the par- 
ticles to draw together from all directions, a given 
volume of cement might always come to the same final 

« 

equilibrium volume, irrespective of the initial propor- 
tions of cement and water. The fact that pastes of dif- 
ferent initial concentrations settle, instead, to dif- 
ferent final volumes per unit of cement was thought to be 
a consequence of the fact that gravity tends simply to 
move the particles in one direction— downward. It was, 
in fact, assumed that lateral concentration of the set- 
tling particles might be completely negligible. These 
assumptions led to the following hypothesis: If a cube 
of cement-water paste occupying only the minimum, or 
base, volume is expanded into a larger cube by mixing 
with additional water, the resultant paste will settle 
until its height is the same as that of the original 
cube. 



On the basis of this hypothesis, Equation 30 of 
Bulletin 2 was developed, in terms of bleeding capacity 
and "excess" water. As shown in Appendix D, the corre- 
sponding relationship between the bleeding capacity and 
the proportion of cement in the initial paste is 

- c B (l - AH') 3 (6) 



-16- 



where c B is the volume of cement per unit of the base 



volume • 



found 



at some paste concentrations will be shown shortly. 
It is now believed that the original theory does not 
attribute sufficient significance to the flocculated 
state of the paste. The role of flocculation is con- 
sidered below. 

Work with emery powders (15) has indicated that 
there is no effective hinderance to lateral movement of 
particles that are not flocculated; in the non-floccu- 
lated state the emery particles always settled to the 
same density of sediment irrespective of the initial 
concentration. Flocculated emery gave much less dense 
sediments and ones which varied in density with the 
initial paste concentration. Since cement pastes are 
flocculated, it also seems probable that for such pastes 
the flocculation is a primary cause of the differences 
in density of sediment that are obtained when different 
initial water contents are used. On the basis of these 
facts alone, the flocculation might simply be regarded 
as the cause of the lack of lateral consolidation which 
the original theory assumes. However, experiments show 
that flocculation also reduces vertical movement. In- 
deed, this can be seen from data presented in Bulletin 
2. It was shown there (p. 72) that by varying the con- 
centration of the flocculating agent (though still main- 
taining flocculation) a large change in the bleeding 
capacity of a suspension of silica powder can be pro- 
duced even though the initial concentration of silica 
is maintained constant. 

To take cognizance of this fact while maintainin 
the original theory it is necessary to assume that the 
size of the base volume is dependent upon the strength 
of the flocculating forces. Evidently, the original 
flocculated structure must be assumed to yield rather 
easily under the action of gravity up to a certain point 
and then to yield no more even though a more compact 



-17- 



arrangement could be produced if the flocculating forces 
were absent. This, in itself, is indeed probable, for 
it seems to be required by other considerations also, 
such as the fact that the bleeding capacity is prac- 
tically independent of the original height of the paste. 
Up to this point the discussion leaves the original 
theory intact, but it shows that the flocculated state 
of the paste is important and must be taken into con- 
sideration. To some extent the influence of floccu- 
lation seems to fit in with the theory of Bulletin 2 f 
but other aspects of flocculation discussed below sug- 
est that the original theory is too simple. 



At concentrations at which there is no channeling 



there is evidently insufficient water to permit the floe 
to form at its loosest possible texture; apparently, 
therefore, a network forms throughout the whole space. 
The cement particles, which make up this structure, are 
presumably held close* to their nearest neighbors (in the 
filaments of the network) by the forces of flocculation, 
so that there could apparently be but little settlement 
if some parts of the structure did not move sideways. 
Perhaps the readjustments are such that lateral movements 
are of secondary importance relative to vertical move- 
ments, but the flocculated nature of the paste makes it 
seem unlikely that sideward movement can be neglected 
entirely. This viewpoint, together with the fact that 
Equation 6 strictly represents the theory of purely 
vertical motion only for certain initial arrangements 
of the particles,** has led to a more empirical study 
of the data. 



quilibrium 



films 



su muse Tioge-Dner as to appear to be in actual contact 
when observed under the microscope. 
**A "staggered" arrangement of the particles in Fig. 16 
of Bulletin 2 would lead to a significantly differ- 
ent result. 



-18- 



Empirical Relationships 

The direct experimental relationship between bleed- 
ing capacity and initial water-content that has been 
found for pastes made from cements of medium fineness 
is shown by Fig. 3. The tests that are represented were 
conducted over a much wider range in water content than 
that originally employed in testing the theory advanced 
in Bulletin 2. The range is also much greater than that 
used in cement practice. This was intentional since a 
primary purpose of the tests was to see whether the 
bleeding capacities for all possible concentrations could 
be represented by one equation. The curves that have 
been drawn through the data-points are parabolic, and be- 
come tangent to the w-axis at the vertices of the parab- 



olas. These curve 



ement 



15341. 



To see how these parabolas compare with Equation 6 
it is convenient to use plots of (1 - AH') 3 versus c in 
which Equation 6 gives straight lines through the origin. 
Such plots are shown in Fig. 4. The parabolic curves of 
Fig. 3 give the S-shaped curves of i?lg. U> which are in 
reasonable agreement with the data. The straight lines 
corresponding to Equation 6 obviously represent only a 
part of the data. Discrepancy at the lower concentra- 



rmai Dieeamg range, a. 
i p.hnrmftlinff occurs, an 



omm 



The parabo lic curves of Fig. 3 become straight lines 
when YAH' is plotted against w. Fig. 5 shows such 
plots both for the cements of Fig. 3 and far a series of 
cements which vary in fineness. The straight lines must 
all pass through the point (1,1) since the bleeding ca- 
pacity, AH*, must equal 1 when w = 1. This aspect is 
of practical value since it means that when the assumed 
relationsnip is applicable the trend line can be es- 
tablished from one good experimental point. However, it 
is evident from Fig. 5 "that cements finer than about 1700 



-19- 



sq cm per gm cannot be relied upon to conform to the as- 



sumed relationship at values of y£H f of 0.2 or less 
(AH 1 s 0.04 or less). 



In contrast with this it has been found that the 
bleeding capacities of nearly all cements, even ones 
having specific surfaces as great as 2^00 sq cm per gn, 
can be repr esented adequately by a straight line on a 
plot of VA H f /c versus w/c if values of w/c are re- 
stricted to ones below 2.6* (w below 0.72). Since this 
limit corresponds to almost 9«5 gallons of water per sack 
of cement (a water-cement ratio of 0.81 by weight), the 
practical range is covered. Fig. 6 shows that this meth- 
od of plotting is satisf actory for all cements of Fig, 
5« The plot of VA H f /c versus w/c has been tried with 
26 cements of medium fineness (1600-2000 sq cm per ©n), 
6 cements of lower fineness, and 8 cements of higher 
fineness (above 2200 sq cm per gm). Only two cements, 
ones having specific surfaces above 2500 sq cm per gm, 
gave data that could not be represented adequately by 
the linear relationship. For these, the bleeding ca- 
pacities were relatively too high at values of w/c of 1.5 
and less, if the line was drawn to represent the other 
data. 



* 



At values of w/c above 3.5 a closely linear relation- 
snip exists between the first powers of AH f /c and 
w/c. The complete curve or AH f /c versus w'/c nas 
some aspects of a hyperooia, but below w/c « 1.5 the 
change in slope is more gradual than the hyperbolic 
curve fitting the upper points. It is of incidental 
interest that if the data are plotted as AH f versus 
w/c, as was done for Fig. 17 of Bulletin 2, they can 
be represented very well by a straight line over a 
considerable range in w/c, but at values of w/c above 
3.5 this relationship fails. It also does not ade- 
quately represent the data at the lowest values of 
w/c. However, the linear relationship is satis- 
factory over a much wider range in w/c than are the 
curves of Fig. 17 of Bulletin 2. 



-20- 





Ct4 



W 



Fig. 3 - Bleeding Capacity versus 
for Cement Pastes erf All Possible 



Water Content 
Concentrations 



The curves shown are parabolic 





Fig. 4 - Comparison of Different Ways of 
Representing Bleeding Capacity Data 

The straight lines represent Equation 6. The S-shaped curves 
correspond to the parabolic curves of Figure 3. 



X 
> 




10 0. 





x 
< 




Fig. 5- Plots in Which the Data Conform to Straight Lines if the 
Relationship Between Bleeding Capacity and Water Content Is Parabolic 

Van' versus w 



2.0 



\£ 



1.6 



1.4 



!» 



Cement Spec Surfoc e 

No. jy cm. p er gm. 

MS02 1720 




— 




























































Cement Spec Surfoc 


e 














/5J4/ /665 










/ 


o 


















































C— 
























c 


s 






















j 


/ 


/ 


























/ 





















































06 0,8 L0 L2 



L4 16 1.8 20 2.2 ZA 2.6 23 ID 
1,0 



Q6 08 10 2 \A L6 



2j0 2.2 24 26 28 10 32 



w /< 



09 



QB 



0.7 



06 



<j 




0L5 - 



04 



03 



02 



01 

















































i 
























































°Z 


/ i 


— — 




_ Cement SpecSurface 

No. s<f. cm per gm 






















S8R-tS6lO iS 7$ f | 












S8R-/S6H 1820 


















Si 


BRi 


562 


z 


22 


00 

(3 


















1 


t ■ ■ -i 


I 


f ' 


/° A 




j 


* 


■ 1 


1 












M 


JA 
























i 


/ 

U 


/ 





















0.2 04 0.6 06 L0 12 1.4 L6 Lfl 10 VI ZA Z6 

Fig. 6 - A Method of Representing Bleeding Capacity Data Linearly 
That Is Applicable to Cements of Practically All Finenesses within 

the Practical Range of Paste Concentrations 

V&p* versus w/ c 



For practical purposes of interpolation and moderate 
extrap olation linear representation of the data on a plot 
of VAH'/c versus w/c appears to be most generally sat- 
isfactory. Two well established points are needed to 
establish the line and more are preferable. However, 
for cements having specific surfaces from 1700 to 2000 
sq cm per gm the slope is usually 0.50 +_ 0.03 and for 
such cements a common value of 0,50 can be assumed if 
correction for only a small change in w/c is wanted and 
data are available for just one value of w/c. 

The point where the straight line cuts the w/c-axis 
is the minimum value of w/c, at which the bleeding ca- 
pacity apparently becomes zero. If this minimum value 
is represented by (w/c) m and the slope of the line by k, 
the equation of the line, as solved for AH f , is 

AH' = k 2 c[w/c - (w/c) m f (7) 

The limit (w/c) m plays the same part in Enuation 7 as 
the eg in Equation 6 except that the two quantities rep- 
resent different ways of defining the indicated limit- 
ing concentration; the additional significance that 
was attributed to c B was wholly hypothetical. As long 
as the limiting, or base, composition is regarded only 
as the practical limit at which no bleeding occurs, no 
;ument can be raised against the validity of the con- 
cept. However, the indicated limit will agree with the 
actual one only if the equation remains applicable at 
the highest concentrations. Equation 7 is believed to 
be sufficiently correct to give good relative values 
of the limiting w/c f s, values at which the actual bleed- 
ing capacities, if not zero, are at least very small. 
Table 3 gives the values of k and of (w/c) m that have 
been determined for cements of medium fineness. For ef- 
fects of fineness see Fig. 6. 

Implications of Bnpirical Relationships 

Alt hough th e assumption of a linear relationship be- 
tween V^H'/c and w/c for values of w/c below 3 appears 



21 



best for general purposes, the linearity of the rela- 
tionship between VSTT" and w (Fig. 5) is really very 
good for the coarser cements. Since for such cements 
the latter relationship is linear over the whole range 
in concentration, it may eventually prove useful as a 
guide to the basic principles that determine the bleed- 
ing capacity. Although empirical, it must at least be 
a close apnroximation to the general law governing the 



bleeding capacities of the cements in question. As 
shown in 'Appendix D, it indicates that the settled volume 
per unit of cement increases linearly with the initial 
water-content, w. It also indicates that the fraction 
of the excess* water in the initial paste that remains 
in the sediment is numerically the same as the fraction 
of the initial paste volume that can be regarded as con- 
stituting base volume. Perhaps this will be clearer if 



written as an equation, namely, 

excess water in sediment base volum 



excess water in initial paste initial paste volume 



.(8) 



The right-hand ratio of this proportion is a measure of 
the concentration and hence presumably of the stability 

of the initial f loc-structure. The equation shows that 
this ratio gives, directly, the proportion of the excess 
water that will be retained by the cements that exhibit 
the linear relationship between V A H f and w. 

In conclusion of this section, Fig. 7 is presented 
to show the different bleeding capacities likely to be 
found for cements of normal fineness in pastes of various 
water contents. Some discussion of the causes of the 
different values found for different cements will be 
given in subsequent sections. 



By excess water is meant the water in the initial paste 
in excess of that needed to form "base" volume with 
the amount of cement that is present. The base volume 
referred to here is, of course, that indicated by a 
plot like Fig. 5. 



-22- 




< 



u 

CO 
Q. 

CD 

(J 



0) 
CD 




1.2 1.4 1.6 

Water-Cement Ratio 



1.8 2.0 2.2 

by Absolute Volume, w / c 



Fig. 7- Ranges in Bleeding Capacities 

Most cements with specific surfaces of 1850 1 100 
sq. cm. per gm. will give pastes whose bleeding 
capacities fall within the band shown. The 

different amounts of bleeding found for different 
cements are attributable to differences in 

chemical constitution and in' treatment as 
well as to differences in fineness. 



Table 3 





Data on 


VAH'/o 


versus w/c 






for Cements of Medium Fineness 








Spec, 


Intercept 


Slope k: 


Cement 




Surf. 


of 




No. 




(Wagner 


w/c -axis t 


aVah'/c 






Turb.) 


(w/c) 


A w/c 






cmTjgm 


m 


w 


LTS-11 




1820 


O.63 


0.51 


LTS-12 




1815 


O.63 


0.50 


LTS-13 




I665 


0.68 


0.50 


LTS-H 1st 


tests 


1880 


0.74 


0.46 


LTS-14 lat< 


2r tests 


1880 


0.66 


0.49 


LTS-15 




1895 


O.63 


0.50 


LTS-16 




1785 


O.58 


0.50 


LTS-17 




1770 


O.76 


0.48 


LTS-18 1st 


tests 


1750 


0.60 


0o50 


LTS-18 later tests 


1750 


0.72 


. 49 


LTS-21 




1630 


0.59 


0.59 


LTS-22 




1775 


0.80 


0.53 


LTS-23 




1875 


0.65 


0.53 


LTS-24 




1925 


0.71 


O.48 


LTS-25 




1825 


0.59 


0.48 


LTS-41 




1915 


0.67 


0.45 


LTS-42 




1920 


0.52 


0.52 


LTS-43 




1965 


0.76 


0.47 


LTS-43A 




1915 


0.55 


0.38* 


LTS-51 




2025 


0.55 


0.45 


SBR-15365 




I665 


0.64 


0.52 


SBR-15496 




1740 


0.73 


0.53 


SBR- 15621 




1820 


0.50 


0.49 


SBR-15b68 




1830 


O.52 


0.52 


SBR-15761 




1800 


O.65 


0.53 



The cement which gave this unusually low value was 
ground from clinker that was not the normal plant 
product • 



-23- 



4. Relationship between Cement Fineness 

and the Bleeding Characteristics 

Bleeding Rate 

Equation 2 indicates that the bleeding rate is in- 
versely proportional to the square of the specific sur- 
face of the cement if other quantities in the equation 
remain unchanged • However, the wi term generally in- 



creases with the fineness of cements prepared from a 
given clinker, as was indicated in Bulletin 2 (p. 44) • 
Data obtained since then are shown in Table 4» Of the 5 
clinkers tested, one fails to show the relationship; but 
there is reason to believe that, for some of the cements 
prepared from that clinker, dehydration of gypsum during 
milling occurred and affected the bleeding character- 
istics* Presumably, if fineness were strictly the only 
independent variable, an increase in it would always mean 
an increase in w i# However, as Table 4 shows, the degree 
of the effect is different for different clinkers. An 
intermediate case is illustrated by Fig. 8. This curve 
can be used to obtain a rough correction for w^ in other 
cases. 



bleeding rates have been accurately determined 



ements made 



un 



1/3 1 

Q versus — — can be used for purposes of in- 

terpolation. The relationship has been found to be fairly 
linear, especially for difference in fineness of only 300 
or 400 sq cm per ©n. For SB* clinker No. 1 the three fine- 



nesses of 10^0, 1665, and 2280 sq cm per gm give an al- 
most exactly straight line. 



Changes in bleeding rate caused by changes in fine- 
ness, and also ones caused by changes in water content, 
can be estimated directly from a plot like Fig. 24 of 

Bulletin 2. which has been reproduced here ns Fix?. Q 



-2k- 



0.30 



0.29 



0.28 



0.27 



0.26 



* — 



0.25 



0.24 



0.23 



0.22 



0.2 



0.20 















































































^0^ 






































■ 






. 


— - _j 














































- 













1200 



1400 



1600 



800 



2000 



2200 



2400 



2600 



Specific Surface, sq. cm. per gm. 



Fig. 8— The Change in w-, with Change 
in Specific Surface as Indicated by 

Data for Cements from Clinker No. 1 



This plot can be used to estimate the effect 
of fineness on the wj values of other cements 
if the changes in fineness are small. 



320 



300 



280 



U 

V) 

d) 

Q. 

■ 

E 



260 



240 



220 



o 



200 



tn 



sO 



c9 

o 

' - X 



80 



60 



E 



a 



40 



(0 



120 



en 

c 

<D 
CO 



00 



80 



60 



40 



20 































- 
























y<0.65<7 = 


--W 


























\a6Z5\ 






























Tk 




*■ 


























\p.6 


00 ^ 




























\a. 


575\ 




















m 




































L 


\o.S5o\ 




























sV.S3S\ 




























\as 


:oo\ 










^h 










^ 
^ 




\p.475 
\p.450\ 




























Tk 


























0.42£\ 














^^w 














^0400 








































■w 















































1000 



500 



2000 



2500 



Specific Surface, sq. cm. per gm. 



Fig.9- Effects of Fineness of Cement and of Water Content 
of Paste on Bleeding Rate, as Indicated by Data for a 

Particular Set of Cements 



Table 4 



Effect of Specific Surface on the Magnitude of w 



■i 



Specific Surface, 
sq cm/gm 



C ement < 
No. 


blinker 
No. 


(A.S.T.V. 
Turb.) 


tfith 
Correction 
for +60 


w i 








micron 
Particles 




- 

SBR-15364 

SBR-15365 
SBR-15366 


1 


980 
1640 
2270 


1040 

1665 
2280 


0.234 
0,269 
0.296 


SBR-15495 
SBR-15496 

SBR-I5497 


2 


1440 

1695 
2500 


1470 

1740 
2500 


0.249 

0.274 
0.320 


SBR-15620 
SBR-15621 
SBR-15622 


3 


1200 

1825 
2290 


1375 
1820 

2200 


0.232 

0.249 
0.266 


SBR-15667 
SBR-15762 

SBR-15763 

SBR-15669 


• 


1200 

1475 
1800 

2290 


1250 
1500 

1800 
2290 


0.249 
0.229 
0.226 

0.249 


SBR-15696 
SBR-15760 

SBR-15761 

SBR -15698 


5 


1260 

1465 
1780 

2255 


1305 
I5OO 

1800 
2265 


0.248 

0.264 
0.285 
0.292 



25- 



(with a change in the bottom scale to units of sq cm per 
©il). Though such a plot is strictly applicable only to 
the cements from the one particular clinker on which it 
is based, it can be used to obtain approximate estimates 
for other cements when their bleeding rates are known 

for one fineness and water content. When the bleeding 
rate for another fineness is sought, the first step is 
to locate the point corresponding to the known bleeding 
rate and the fineness at which it was obtained. Then, 
regardless of whether the water-content indicated by 
the plot is the same as was used in the test, it is 
treated as if it were, and thus the point corresponding 
to the new fineness is found. The bleeding rate indi- 
cated by this point is approximately that which would 
be obtained experimentally at the new fineness. If the 
effect of a change in water content is sought, the pro- 
cedure consists first in locating the point correspond- 
ing to the experimental bleeding rate and the water- 
content at which it was obtained, then in following 
along the vertical (constant specific surface) line 
through the initial point until the point corresponding 
to the new water content is reached. The bleeding rate 
indicated by this point is the estimated value for the 
new water content. These methods are simply convenient 
means of obtaining approximate estimates, but theyusu- 
ally check with experiment rather well. 

Bleeding Capacity 

The effect of cement fineness on the bleeding capa- 
city is shown by the data of Table 5. These data indi- 
cate that the decrease in bleeding capacity for an in- 
crease of 100 sq cm per gm in the specific surface of 
the cement is usually 0.011 _+ 0.004 within the practical 
range of concentrations. The data vary for the dif- 
ferent groups of cements j when more than a small change 
in fineness is involved, the resultant bleeding capacity 
had best be determined by direct test if an accurate 
value is wanted. 



-26- 



Tabl« 5 

Effect of Cement Fineness on Bleeding 

Capacity f AH' 



At w/c-1.6 



At 



Cement Clink- 
No. er 
No. 



SBR 
SBR 
SBR 

SBR 
SBR 
SBR 



15364 
15365 
15366 

15495 
•15496 

15497 



SBR-15620 
SHR-15621 
SHR -15622 

SBR -15667 
SBR-15762 

SBR- 1576 3 
SBR-15669 

SBR-15696 
SBR-15760 
SBR -15761 
SBR-15696 



Sp.Surf 

£ t 

cm / £pi 
(Turb.) 



1040 

1665 
2280 

1470 

1740 
2500 

1375 
1820 

2200 

1250 
1500 
1800 
2290 

1305 
1500 

1800 

2265 



Bleeding Change in 
Capacity, &H 1 per 

AH' 100 

cm /gm 



w/c - 



1.8 



0,174 
0.093 
0.027 

0.119 
0.074 
0.014 

0.154 
0.106 

0.054 

0.167 
0.147 
0.126 

0.083 

0.145 
0.126 

0.09- 

0.061 



0.0130 
0.0107 

0.0157 
0.0079 

0.0108 
0.013? 



0.0080 

0.0070 
0.0088 

0.0097 
0.0106 

0.0071 



Bleeding Change in 
Capacity t z^H 1 per 

AH ■ 100 



At w/c - 2.0 



cm 



/m 



0.219 

0.133 
0.051 

0.172 

0.114 
0.028 

0.202 
0.150 
0.091 

0.214 
0.194 
0,171 
0.124 

0.192 

0.171 

0.137 
0.100 



hange 



cm e /gm (omitting SBR -15495) 

0.010 



0.0138 
0.0133 

0.0215 
0.0113 

0.0117 
0.0153 



0.0080 
0.0077 

0.0096 

0.0108 

0.0113 
0.0080 



Bleeding Change in 
Capacity, ^H' per 
AH • 100 

cm /gm 



0.256 
0.169 

0.079 

0.206 
0.146 
0,041 

0.237 
0.185 
0.123 

0.249 
0.228 
0.206 
0.156 

0.225 
. 20 5 
0.170 
0.130 



0.0139 
0.0146 

0.0222 
0,0139 

0.0117 
0.0163 



0.0084 
0.0071 
0.0102 

0.0103 

0.0117 
0.0086 



0.011 



0.0115 



-27- 



5» Relationships between Bleeding Rates 

and Bleeding Capacities 

The Experimental Evidence 

A comparison of Figs. 2 and 7 shows that bleeding 
rate and bleeding capacity increase in similar manner 
when the water-cement ratio is increased. Indeed, for 
a given cement, the bleeding rates are almost propor- 
tional to the bleeding capacities, at least for bleeding 
rates up to 300 x 10 "6 cm per sec. However, the straight 
line that best represents the data on a plot of bleeding 
rate, Q, versus bleeding capacity, AH*, commonly cuts 



the Q-axis at a small positive value, generally less 
than 30 x 10-6 cm per sec. (See Fig. 10) At values of 



Q above 300-350 x 10-6 cm per sec the data-points tend 

to fall above the straight line that represents the lower 
values.. 



The relationship between bleeding rate and bleeding 
capacity sometimes remains quantitatively unchanged as 
the fineness to which a cement is ground is changed 
markedly. This is most common within the range of moder- 
ate to high finenesses; a cement coarser than 1500 sq 
cm per gm is likely to show a higher bleeding rate for 
a given bleeding capacity. Fig. 10 shows the relation- 
ships between bleeding rate and bleeding capacity for 
cements from five clinkers. For only one clinker does 
the slope of the data-line decrease significantly with 
each increase in fineness throughout the whole range 
from coarse to very fine. This is for clinker No. 5 
which is the only one that is high in alkali (total al- 
kali oxides = 1.57ft That this alkali is probably partly 
responsible for the decreases in slope is indicated by 
data to be discussed in the next paragraph; the rate at 
which the alkali goes into solution increases, of course, 
with the fineness of the cement. 

It is apparent from Fig. 10 that the relationship 
between bleeding rate and bleeding capacity differs some- 



ame 



• II 



-28- 



020 




a20 



Bleeding Capacity, AH' 



Bleeding Capacity, AH 




aos 0.10 015 

Bleeding Capacity, AH* 



a to 



Fig. 10 



Relationships Between Bleeding Rates 
and Bleeding Capacities 




0.05 



0.10 

Bleeding 



0.15 0.20 

Capacity, A H 



0.25 



Fig- 



A Composite Plot of Bleeding 



Rates versus Bleeding Capacities for 
Cements of the LT5 Group 



different clinkers. In Fig. 11 the data for the cements 
of the LTS group that are not high-early-strength or 
air-entraining cements have been placed on one plot to 
show the degree of spreading. Data for the cements 
that showed free alkali in excess of 6.8 by the Merriman 
test have been represented by solid circles. There were 
five of these cements and it will be noted that their 
data all fall along the lower edge of the band. Appar- 
ently, alkali in a cement tends to lower the ratio of 
bleeding rate to bleeding capacity, as was inferred 
from Fig. 10. The two straight lines in Fig. 11 indicate 
the approximate limits of the range in the data when 
the five high-alkali cements just discussed, and two 
cements (Nos. 13 and 22), which show the highest ratios 
of Q to AH 1 , are excluded. (The plot was limited to 
values of Q, below 350 x 10"° cm per sec.) 

To indicate the degree of correlation between bleed- 
ing rates and bleeding capacities when the data are for 
one common water content and have been adjusted to repre- 
sent one fineness of cement, Table 6 is presented. The 
cements are arranged in the order of their bleeding rates 
but they will be seen to be also lined up roughly in the 
order of their bleeding capacities. The significance 
of the coefficient of correlation is discussed in sec- 
tion 7. 

From what has preceded it is evident that in spite 
of differences in the finenesses and chemical composi- 
tions of the cements, and in the water contents of the 
pastes, if the bleeding rate of one paste is found to 
be considerably greater than that of another, it is prob- 
able that the bleeding capacity is greater also. More 
knowledge of the cements can change this judgment in 
any particular case but, as a rule, it holds more often 
than not. Reasons for this, as well as for the departures 
from the general relationship, are given in the follow- 
ing discussion. 



-29- 



Table 6 



B leed ing Rates and Bleeding Capacities as 
Adjusted to Correspond to a Specific Surface 

(Turb.) of I84O sq cm per gm 

From tests at 23.5°C using a water- 
cement ratio of O.466 by weight. 





Bleeding Rate, 


Bleeding 


Cement 


Q, cm per sec, 


Capacity, 


No, 


x 100 


AH' 


LTS-A2 


163 


0.123 


LTS-23 


156 


0.096 


LTS-51 


144 


0.097 


LTS-15 


HO 


0.084 


LTS-21 


140 


0.099 


LTS-16 


129 


0.097 


LTS-11 


117 


0.078 


LTS-12 


116 


0.081 


LTS-18 


115 


0.068 


LTS-22 


111 


O.O46 


LTS-25 


107 


O.O84 


LTS-13 


104 


0.048 


LTS-43A 


98 


0.061 


LTS-43 


88 


0.067 


LTS-41 


85 


0.069 


LTS-24 


80 


0.062 


LTS-H 


79 


O.O54 


LTS-17 


78 


0.050 


Coefficient < 


>f correlation between 




bleeding ] 


rate and bleeding capacity: 


0.84 


Odds against 


chance correlation: 12,30C 


1 to 1 



-30- 



Theor et ic al - C ons i derat i ons 

* 

That both bleeding rate and bleeding capacity should 
increase as the dilution of a paste is increased is easy 
to understand and needs no elaboration. It can also 
be readily seen that an increase in fineness should re- 
duce the bleeding rate since it reduces the sizes of 
the capillary spaces through which the liquid flows be- 
tween particles. -That bleeding capacity should also be 
reduced is not so immediately evident.- Indeed, a given 
volume of inert particles of a given shape and a given 
relative grading would give a constant gross volume of 
sediment, irrespective of fineness, if a given arrange- 
ment of particles were maintained. The fact that the 
bleeding capacity for a given water-content does ordi- 
narily decrease with fineness can be attributed to the 
flocculated state of the paste and to the fact that a 
little chemical reaction evidently occurs during mixing. 
With a constant flocculating force per unit of surface, 
the greater ratio of surface to particle-weight in a 
finely ground powder should cause the powder to stop 
settling before it had developed -as compact an arrange- 
ment as would be reached by a coarser powder. This could 
account for the whole of the observed effect, but prob- 
ably chemical reaction plays a part also. As already 
stated, a coating of hydrate is believed to be formed 
on the cement particles when they first contact the water. 
Hence, for a given thickness of layer the total solid 
volume is greater the finer the cement, and can have 
re of an effect on the bulk of sediment. 

Differences in the chemical reactivities of dif- 
rent cements may be expected to affect bleeding rate 



and bleeding capacity somewhat similarly if they are 
manifested principally -thro ugh different thicknesses 
of reaction layers on the cement particles. The relative 
thicknesses of the layers would determine the relative 
spaces between particles, and hence the relative bleeding 
rates. They would determine the relative effective di- 
ameters of the particles and hence also the relative 
volumes of sediment, provided the arrangements of the 



-31- 



particles were the same. This effect on the volumes of 
sediment would, presumably, be too small to account for 
all the observed differences, but loosely formed hydrate 
layers of different thicknesses would be expected to pro- 
duce differences in the structures of the settled pastes. 
Such layers could deform at points of contact as the par- 
tides came to rest, and they could thus broaden and 
strengthen the contacts. Thicker layers would broaden 
the contacts more effectively than thinner layers and 
could thus stabilize less compact structures. 



The speculative and merely qualitative elements in 
matters discussed constitute one reason why caution 
.eeded in estimating bleeding capacities from bleed- 



ing rates. A more positive reason to expect departures 
from good correlation arises from the possibility that 
there may be material differences in the strengths of 
the flocculating forces developed by different cements. 
When the flocculating forces are strong enough to main- 
tain the floe structure intact, an increase in the 
strength of the flocculating agent has little effect on 
the bleeding rate; this is shown by the tasts on silica 
powder that were reported in Bulletin 2. The same tests 
show, however, that the bleeding capacity is strongly 
affected by the strength of the flocculating agent. 
Hence, if two cements are otherwise the same but de- 
velop flocculating forces of materially different 
strengths they may show the same initial bleeding rates 
but distinctly different bleeding capacities. The fact 
that there is as much correlation between bleeding rates 
and bleeding capacities as is found when different ce- 
ments are involved seems to indicate that usually the 
strengths of the flocculating forces are not widely dif- 
ferent for different cements. Possible differences in 
flocculation will be considered again when the effects 
of alkali hydroxides are discussed. 



-32- 



6. Duration of Bleeding 

For heights of paste -sample for which bleeding is 
completed before the end of the so-called dormant pe- 
riod, the duration of bleeding is often strikingly in- 
sensitive to change in the water content of the paste 
and in the fineness of the cement. Bleeding times for 
a constant height of sample (3.3 cm) are given in Table 
7. Many of these data show the constancy just mentioned. 
However, for the very finely ground cements the bleeding 
time increases with the water-content of the paste. 




In Table 7, the figures for * i QQ — are the 

times, in minutes, in which bleeding would have been 
completed had the settlements continued at the initial 
rates. The ratios of the actual bleeding times to these 
"minimum" times are given in the columns headed R. 
Though apparently not closely reproducible, these ratios 
are, in all cases, practically independent of the water- 
contents of the pastes. They average about 1.5 for most 
of the cements but are significantly higher in a few in- 



stances. Since R is so independent of wat er -c ont ent , 
maintenance of a practically constant bleeding time when 
the water-content is changed usually means that the ratio 
of AH f to Q, has stayed fairly constant. 

For pastes that complete their settlement within 
the dormant period the bleeding time is approximately 
proportional to height of sample. Though this rela- 
tionship is only approximate (Bulletin 2, p. 8), it seems 
to hold fairly well in most cases, when wall effects are 
avoided. Fig. 12A shows the curve for a cement that had 
an estimated dormant period of 60 minutes. Tests, in 
the same vessel, on cements with longer dormant periods 
have usually shown considerable departure from the 
straight line before the Bnd of the estimated dormant 
period (see Fig. 12B) , but this may be because the 
heights of the samples considerably exceeded the diam- 
eter of the vessel (see Bulletin 2, p. 141). When 
bleeding time is proportional to paste height, the ratio 
(R) of the total bleeding time to the time based on the 



33 



Table 7 

Bleeding Times and Other Data 
for Cement Pastes 3*3 cm In Height* 

R - Ratio of bleeding time to ?*? ^ H ' 

60Q 



Clink- 










Count 












er 




Coarse 




Med 


Lima 
•15365 






Very 

SBR- 


Fine 
15366 




Ref. 




SBR-. 


15364 




SHR- 




No. 


w 


Time 


. 3.3 ^H' 


- R 


Time 


>, 3-3 ^fl 


f 

R 




Tlme 


. 3.3 AE 


1 
- R 






nda. 


60Q 


A\ 


* aln. 


60Q 


AW 


w 


sin. 


60<J 




0.457 


34 


22 


1.55 


0.518 31 


23 


1.35 


0.520 


20 


14 


1.4 




0.478 


28 


21 


1.3 


0.568 33 


24 


1.4 


0.570 


26 


19 


1.35 


1 


0.498 


33 


22 


1.5 


0.618 46 


• 30 


1.5 


0.620 


29 


23 


1.25 


0.517 


32 


21 


1.5 


0.670 43 


30 


1.4 


" O.65O 


39 


27 


1.45 




0.517 


28 


19 


1.45 








0.720 


39 


30 


1.3 




0.543 
0.490 


32 


22 


1.45 






• 












SBR- 

39 


■15495 
28 




SBR- 


15496 






SBR- 


-1 5497 
14 






1.4 


0. 516 33 


20 


1.65 


0.568 


20 


1.45 


2 


0.514 


U 


27 


I.65 


0.566 37 


24 


1.55 


0.618 


25 


18 


1.4 




0.539 


39 


28 


1.4 


0.617 45 


28 


1.6 


0.668 


31 


23 


1.35 




0,564 


42 


28 


1.5 


O.665 44 


30 

15621 


1.45 


0,718 


38 


29 


1.3 






SBR-15620 




SBR- 






SBR- 


■15622 






0.470 


50 


29 


1.7 


0,501 52 


34 


1.55 


0.542 


44 


27 


1.65 


3 


0.495 


55 


34 


1.6 


O.542 48 


34 


1.4 


0, 562 


45 


29 


1.55 


0.520 


48 


28 


1.7 


0.581 50 


38 


1.3 


0.601 


56 


36 


1.55 




0.545 


53 


29 


1.85 


0.622 52 


37 


1.4 


0.628 


55 


36 


1.55 












0.671 50 


. 36 


1.4 


0.649 


55 


37 


1.5 










.. 








0.702 


52 


32 


1.6 




0.510 


SBR- 


-15762 




SBR- 


15763 








50 


35 


1.45 


0.550 57 


36 


1.6 










L 


0.540 


60 


31 


1.95 


0.575 64 


36 


1.8 










^P 


O.56O 


60 


33 


1.8 


0.600 60 


35 


1.7 












0.600 


60 
SBR 

52 


31 
-15760 


1.95 


0.625 60 

SHR- 


28 
15^61 

34 


2.15 






m /~ / ■*■ — r 






0.550 


■*■• 1 v " 
30 


I.75 


0.550 58 


1.7 


0.550 


^BR- 
50 


15693 

33 


1-5 
1.4 


5 


0.600 


53 


36 


1.45 


0.600 58 


39 


1.5 


0.650 


68 


49 




0.625 
O.65O 

•The a 

3 A 


52 

51 


31 
29 


1.7 
1.75 


0.600 60 
O.65O 60 


43 

40 


1.4 
1.5 


0.700 


75 


48 


1.55 




ctual 


jpaste heights 1 


rere not all 


3.3 cm, 


but the 


bleeding times were 






adjusted by assxaalng blei 


>ding tine to be proportional 


to past 


e height. 





-34- 



80 



70 



60 



50 



40 



30 



20 



10 













Jk 






































(h 


























\T 










°J 
























































J 










Cement SBR-I563I 














A 


Dormant period: 60 mm. 
Dia. of test iar : 100 mm. _ 














| 


i 


i 


i 


1 







90 



CD 80 



CD 



40 



10 



10 









Cement SBR-IS666 
Dormant period'. 115 min. 
Dia. of test jar : 100 mm. 



10 20 30 40 50 60 70 60 90 fOO 110 

nitial Height of Paste, mm. 



120 130 140 



Fig. 12- Bleeding Time versus Height of Paste-Sample 



initial rate is about the same for different heights. 
This may be seen fron the following equation, since both 
Q and AH* are practically independent of paste height 
under proper test conditions. , 

AH» /initial 
Total bleeding time, = R x j > . — x f height 

in min. I of paste 

\ in cm 



7« The Dominant Chemical Influences 



■■ 



in Cements of Normal Production 



As stated near the end of section 2 it is now be- 
lieved that the cement particles develop coatings of hy- 
drous reaction products as soon as they come into contact 
with the mix water. To learn more about the dominant 
chemical influences, the bleeding data now on hand for 
cements of ordinary fineness* have been studied in re- 
lation to the other properties of the cement3. The ce- 
ments had been submitted to unusually extensive and 
varied analysis and testing; hence, many kinds of data 
were available for this comparison. However, only a few 
of the properties showed good correlation with the bleed- 
ing. These will be discussed in this section. 



Statistical Measures 



The "scatter" of the points in the plots that are to 
be presented is very great as compared with results ob- 
tained in the usual kind of physical .studies in which 
the aim is to have only one independent variable. In 
the present studies since cement composition varies in 
a random way, the scatter is to be expected. The purpose 
here is not to establish exact relationships between the 
variables but simply to show whether correlations are 



*To minimize the effects of fineness, only cements hav- 
ing specific surfaces between 1625 and 2025 cm /gm 
(A.S.T.M.) were utilized in this study. The bleeding 
data were used both as determined and after adjustment 
to correspond to I84.O cm /gjn« 



-35- 



indicated and how good the indications are. To this 
end, coefficients of correlation, and the odds against 
the correlations 1 being purely fortuitous, have been cal- 
culated and have been recorded on the -nlots. 



The coefficients are the Pearson coefficients of 
correlation (12) which are widely used by statisticians. 
They may be defined for present purposes simply as index 
numbers which indicate the degree to which a linear re- 
lationship can be assumed between two variables. The 
value is zero when there is no correlation, and is + 1 
when the points all fall on one straight line. A neg- 
ative sign indicates that as the one variable increases 
the other decreases. It is not easy to state the exact 
significance of any particular value between and 1, 
though of course the greater the absolute magnitude the 
greater is the indicated correlation.* The present in- 
terest, however, lies not so much in the degree of cor- 
relation which the data indicate but in the judgment 
which this makes possible as to whether the two proper- 
ties being studies are really dependent on each other. 



However, the coefficient, or rather 100 times it, may 
be said to represent approximately the percentage of 
"overlapping" between the deviations of the two var- 
iables from their averages. That is, if the coef- 
ficient is 0.60, and the conditions are symmetrical 
with respect to the two variables, 60$ of the dif- 
ference of one variable from its mean is determined, 
on the average, by 60# of the difference of the ot 
variable from its mean, and vice versa, there being 
thus l&% of each departure from the mean that is de- 
termined by factors that are independent of the value 
of the other variable (12). For example, if the two 
variables that are being compared are % and B, but 
there are other variables C, D, E, etc., then if C 
influences A but not B, C is one of the factors deter- 
mining the 1+0% of the variation in A that is not 
attributable to B. Likewise, D might affect B but not 



i-i. # 



-36- 



That is, though a calculated coefficient of correlation 
represents the particular data on which it is based, it 
is only an approximation to the value that would be ob- 
tained if an infinite number of pairs of data were avail- 
able. The latter value can be considered to give the 
true correlation between the two properties that are be- 
ing compared. Hence, the question of most immediate con- 
cern is this : Having oDtained experimentally a given val- 
ue of the coefficient of correlation what are the chan- 
ces that an infinite number of tests would also show cor- 
relation of the same kind, though perhaps different in 
degree? Since the infinite number of tests is not avail- 
able, such a question may appear to be impossible of 
solution. However, since only chances, or odds, are 
under consideration, a solution can be obtained on the 
basis of the theory of probability. As might be expect- 
ed, the result depends not only upon the size of the co- 
efficient that is found experimentally, but also upon the 
number of pairs of data on which the coefficient is 
based, for, obviously, any particular value found for the 
coefficient is more likely to denote a true relationship 
if the number of tests is large than if it is small. It 
is not advisable here to go fully into the method of es- 
timating the odds, as it can be found in standard texts.* 



*The method of calculating odds that was used for this 
report was based on the "normal" probability curve. 
The ratio of the coefficient of correlation to the 
"standard e rror for zero correlation" (which latter is 
simply 1/ aJ N-l , where N is the number of pairs of 
tests) was used as entry to a table of the probability 
integral. By this means the numerical probabilities 
were obtained, both for and against the assumption 
that the correlation was fortuitous. The* ratio of 
these probabilities gives the odds. Details and fur- 
ther explanation are given in reference (12). There 
is a more precise method (12) based on the work of 
Student, but since it gives greater odds the results 
reported here are at least on the conservative side. 
Some authors (12) report the odds against obtaining 



37 






However, the odds may at least be defined more precisely 
than was done on the plots. They are the odds against 
obtaining purely by chance a coefficient of correlation 
as large as that found experimentally. That is, if there 
is in fact no true correlation, these are the odds a- 
gainst obtaining by chance a coefficient at least as 
large numerically (i.e., either + or -) as was found. 

Since the purpose in figuring the odds is to enable 
one to form a judgnent as to whether there is true cor- 
relation, it is worth noting that, in statistical stud- 
ies, odds of less than 20 to 1 (calculated as for this 
report) are often disregarded. On the other hand, odds 
of around £00 to 1 are usually considered highly sig- 
nificant. Odds of 20 to 1 correspond to a probability 
against correlation that amounts to about 5$ of cer- 
tainty, whereas the corresponding probability for odds 
of 400 to 1 is only 0.25$. 

With these facts in mind, attention can be turned 
to the experimental data. For simplicity, all plots have 
been based on the same sets of data for bleeding rate and 
bleeding capacity, respectively, except for Figure 17, 
which is introduced to show that other sets of data 
sometimes give better correlation. In all the other 
plots the bleeding data correspond to a water-cement 
ratio of 0.^66 by weight and a temperature of 23.5°C 
(7A°F). The data have been adjusted to correspond in 
each case to a specific surface of 1&L0 so cm ner 




wi 



appear anc 



hown on the diagrams differ more 
bo be justified by the difference 
.ots. The odds, however, are in 



by chance a correlation having the same sign (+ or -) 
as is indicated by the data. Such odds are about twice 
as great as those calculated for this report, which are 
based only on the absolute value of the coefficient. 
Hence, here again the method that was used gave con- 
servative results. 



-38- 



no sense a linear measure of the "scatter". The coef- 
ficient of correlation is more nearly, thoughnot strict- 
ly, of that nature. The fact that the odds increase so 
rapidly for a small improvement in correlation is not 
something that can he "seen" in the plots, hut is simply 
an outcome of probability theory— ^hich has as one of its 
purposes" the avoidance of errors of purely intuitive 
judgment. 

As a matter of fact, no great significance should 
be attributed to the exact odds when they are very high, 
for the method of calculation is approximate only. The 
main point to observe is simply whether or not the odds 
are considerably greater than a value which one is will- 
ing to accept as probably indicating true correlation. 



Heat Liberation 

A relationship between bleeding capacities and rates 
of heat liberation, both determined at 90°F, was noted 
first by W. C. Hansen. Study of other sets of heat data, 
both rates of heat liberation and cumulative heats for 
periods of 1 hour or less, has shown in many cases rath- 
er good correlation with both bleeding rate and bleed- 
ing capacity. Two of the better correlations are illus- 
trated by Figs. 13A and 13B, which show the rates of 
heat liberation after 1 hour at 75°F plotted against 
bleeding rate and bleeding capacity, respectively. Figs. 
HA and 14B are based on the amounts of heat liberated 
during the first 5 minutes at 75° F « 

The sizes of the coefficients and the odds against 
chance correlation indicate that the rate and amount of 
bleeding are dependent to a considerable degree on some- 
thing that is a prominent source of heat during the 
early history of the paste. Heats of solution contribute 
significantly to the heat liberated during the first few 
minutes, but the good correlation between the bleeding 
and the heat liberation at one hour makes it highly prob- 
able that chemical reaction is responsible for all the 
various correlations with the heat data. Reactions of 



39 



■ 

tricalcium aluminate appear most probable in this con- 
nection because of the relatively great reactivity of 
the C3A and because, in reacting, C3A liberates more heat 
per gram than any other cement compound except free lime. 
Only reaction which occurs during the mixing period can 
affect the initial bleeding rate; the correlation with 
the later heat liberation indicates simply that the same 
reaction is continuing. That the further reaction leaves 
the initial bleeding rate unaffected for a considerable 
period is evidently due in large measure to the fact 
that the reaction rate quickly falls to a small value. 
(See Fig. 1* and also the discussion in section 2.) 

Readily Leached SO3 

Another property which was found to correlate rather 
well with the bleeding is the amount of S0 3 that can be 
leached from the cement in a short time. The concentra- 
tions of S0 3 obtained by Hansen and Pressler (8)** in 
five extractions of each cement were added together to 
obtain the values plotted in Fig. 15. The odds against 



ne total heat liberated during the first 5 minutes is 
usually 3 or U calories per gram of cement, whereas 
that liberated during the next half hour is less than 
a calorie. Heat of solution during the 5-minute pe- 
riod is apparently only a few tenths of a calorie, ex- 



cept ror the effect of free lime. Free lime can in- 
crease the heat of solution materially, but in gen- 
eral the initial heat liberation is judged to be 
sufficiently greater than is to be expected from heat 
of solution to justify the assumptions made regarding 
chemical reaction. 
**A change in the method of extraction placed the data 
for 5 of the cements on a basis somewhat different 
from the rest.. However, one cement was tested by 
both methods and the concentrations checkBd closely. 



-40- 



u: 



(0 



a, 

■ 

£ 

l. 



u az 




12 



LO 



05 



04 



04 



2 

















© 




























A 














^ 




oo 
















\p © 


> 










© 


o 


f\ 


< 














w 





A^o 


* 







Q* 10* (for X- 0.466, at 74*F) 

Fig 13 - Rate of Hear 



200 



ao4 



0.0 6 



aoa 
AH 1 



o-io 

1 For IS 



0.1? 14 ai6 

0.4 66, at 74* F) 



O.ld 



Liberation versus Bleeding Properties 

(23 Cements) 



® Correlation with bleeding rate 
Coefficient of correlation: -0,74 
Odds against chance correlation: 2070 to ! 



Correlation with bleeding capacity 

Coefficient of correlation : -0.72 

Odds against chance correlation- 1380 to I 



• 



c 
6 







100 120 140 160 

Q x 10* (For »-« 0.466. at 74 # F.) 



5.0 










1 ' 


i a 


[\ 
















v! 


p 


40 


< 


















o pv 

© o \ 

b CD^ 


Q 



















e 
















10 




























< 












o 










yo 








zo ] 










o 


* ^^ 













■ -\ 






v 


It 

























200 



004 



0-06 



0.1? 



0.14 



aoe oio 

AH' (For ™ .Q466 ( at 74 # F) 



16 



3 18 



Fig. 14 -Heat Liberated In the First Five Minutes versus Bleeding Properties 

(23 Cements) 

® Correlation with bleeding rate 
Coefficient of correlation : -070 

Odds against chance correlation: 1035 to 



® Correlation with bleeding capacity 

Coefficient of correlation: -0.57 

Odds against chance correlation [2 7 to I 



LB 

u 

i- 

t W 

? 

iZ L0 

c 

■ — 

0-2 






o 






^ 


;\ 


d 











w 


o 


o 


>wO 















1 


>. 


° o 










o o 




r ^v 


o 





60 



SO 



100 



120 140 

Qx 10* 



160 



ISC' 




Fig. 15 



Sulfate in Five Extracts versus Bleeding Properties 

(«8 Cements) 

@ Correlation with bleeding rate 
Coefficient of correlation*. -0 70 
Odds against chance correlation-. 235 to I 



Correlation with bleeding capacity 
Coefficient of correlation: -0.43 
Odds against chance correlation: 12 Yo 1 




u 








1 






[ 




Lfl 














_/R\_ 




IflV 














vB> 








o N 


o o 
o 












14 








3 














o 


B ^^ o 












LO 


















V*P^ 






• 
















o 




^^ 










Q6! 
















HA V 








" n 


\ 


o 






m 




. 






\ 









o.04 o.o6 o.oa ato 0-iz a i4 a>6 

AH* 



Fig. 16 -Sulfate in Five Extracts ♦ 0.05(%C 3 A-7) versus Bleeding Properties 

(18 Cements) 

® Correlation with Dieedlng rate 
Coefficient of correlation : -0.76 
Odds against chance correlation: 550 to 1 



(§) Correlation with bleeding capacity 
Coefficient of correlation : -051 
Odds against chance correlation: 28 to 1 



chance correlation are not as large as for the previous 
plots, but they can be naterially improved by making an 
addition to the sulfate value when a cement is high in 
tricalcium aluminate. This is shown by Fig. 16 in which 
one-twentieth of the C3A in excess of 7% was added to 
the SO3 figures. 



It will be noted that the data do not correlate as 
well with bleeding capacity as with bleeding rate. How- 
ever, better correlation is obtained with other bleeding- 
capacity data, as is shown by Fig. 17. This, together 
with the fact that the correlations are of the same sign 
as those with the bleeding rate, gives reason to believe 
that the relationships are significant. It appears, how- 
ever, that bleeding capacity is less closely related to 
the SO3 concentration than is the bleeding rate. 

The coefficients of correlation with the bleeding 
data are of such magnitude, both for the heat data and 
for the soluble SO3, as to indicate that the correla- 
tions result, in part at least, from a common cause. 
From the preceding discussion it appears that this under- 



lying cause is chemical reaction, very probably reaction 
of tricalcium aluminate. The correlation between the 
bleeding and the SO3 concentrations can be plausibly ex- 
plained on the basis of such reaction, because a prin- 
cipal reaction-product of C3A is calcium sulf oaluminate , 
for which sulfate is needed. The mass-action effect of 
the sulfate in solution may be quite important in deter- 
mining the early rate of this reaction. There may also 
be a less direct connection between the extent of re- 
action and the sulfate in solution; alkalies often ac- 
count for a large part of the soluble sulfate and Lerch 
(11) has concluded that alkalies in the cement probably 
increase the reactivity of the aluminates. The improve- 
ment in correlation that is brought about by weighting 
the SO3 figure when the cement is high in C3A tends of 
course to support the assumption that a reaction of C3A 
is involved. 



-a 



Soluble Alkalies 



If the correlation between bleeding and extracted 
SO3 is dependent in part upon effects of the alkalies, 
correlation between the bleeding and the amounts of dis- 
solved alkalies is to be expected. That there is such 
correlation with respect to bleeding rate is shown by 
Fig. 18. It appears, however, that the correlation is 
not as good as that between the sulfate and the bleeding 
rate; hence, there is room for the opinion expressed 
above that the concentration of sulfate may directly 
affect the rate of formation of the calcium sulfoalu- 
minate. Also, since dehydration of gypsum can super- 
saturate the solution with calcium sulfate, precipita- 
tion of gypsum may sometimes be a factor in reducing 
the bleeding, a factor which would contribute to the 
degree of correlation between the dissolved sulfate and 
the bleeding, but not to that between the alkalies and 
the bleeding. With respect to bleeding capacity, par- 
tial correlation with the alkalies is indicated, but 
the odds against chance correlation are too low to give 
confidence in the relationship. 



arent Increase of the Dens it v of Cement in Contact 




with 7/ater 



One of the properties that correlate well with the 
bleeding is an apparent increase in cement density caused 
by contact with water. The density determined in the 
usual way, in kerosene, is presumably correct • However, 
when water is used instead of kerosene, in a LeChateliei 
flask or other type of pycnometer, the computed cement 
density is higher because interaction between the cement 
and water reduces the total volume. The pycnometer for- 
mula makes this appear as an increase in cement density. 
That this apparent increase in cement density [ deter- 
mined after about 20 minutes of contact between cement 
and water) is significantly related to the bleeding 
properties is shown by Fig. 19. 



-42- 




_ 




Fig. 17 - Relations hi B >n Extra i d fat 



Bleed 



Capac s for £ -2.0, at 74* F *,-0t6 

(ifl ements) 

® Correlation vten sulfate n fivf txtra 

dnd blading capa I j 
C ■ icient r jrn . -0S6 

Oddi aga nee r elation 55 to I 

® Correla b*!wt#n ife a fl 

♦ O05(%CjA dnd bleed | ap j 

Coe' I flf perflation 

Odds tg^ins >rrf - • 



AH 

and 



ua 



azz 



LO 



03 



to 






- Ok 

m 



n 
2 











1 2D 40 

Q < 10* 



reo 



ISO 



flg 




Fig IS -Wa ; er-5oluble Alkali A f versos Bleeding Prope I es 

(1 8 erne i • 

Correla otsedtng rate 

Coei' en* 3f" co<*rel fi 

Oo^5 against cha 



® Correlation witn ble* - -j capacity 

] 5 aqai ?tat»ocv 3 ?o I 



a. 



c 
o 



10 

u 

c 



c 

QJ 

CL 
< 




0.08 



0.07 



0.06 



60 



SO 



too 



100 




\10 140 

Q x I0 6 " H 

Fiq. 19 - Apparent Increase in Density of Cement in Water 

versus Bleeding Properties 

(23 Cements ) 

® Correlation with bleeding rate 
Coefficient of correlation: -0.80 
Odds against chance correlation: 5440 to 1 




Correlation with bleeding capacity 
Coefficient of correlation: -0.S4 
Odds against chance correlation: 92 to 1 



12 

< 

*-» IB 

Si 
§ 

+ L4 








O 




/ 
















r o o 












: 












z 

TO 
u 

"5 LO 

Ul 








o 


^ o 


o 


























> 

c ox 




s 


v^Q O 


o 












if /° \ 














o 

az 


s 


)/ 

















0.02 0X>4 O05 a06 0,07 0.06 0.0* 

Apparent Increase in Density, gms. per cc. 

Fig. 20 - Sulfate in Five Extracts + 0.05 (%C 3 A-7) versus Apparent 

in Density of Cement in Water 

( 16 Cements) 

Coefficient of correlation: -0.86 

Odds against chance correlation: 2700 to I 



ncrease 



From 



£-11 



at endwater used i 
est, the volume changes attending the i 
sulfates can be estimated. It is found 



■cney are surficient to account for the major part of the 
differences between the apparent increases in density of 
the various cements. A reasonable amount of calciumsul- 
foaluminate formation could account for the rest. Ac- 
cording to this explanation the correlation between 
bleeding and the apparent increase in cement density 
follows directly from the correlation shown in Fig. 16. 
This view is strongly supported by the fact that cor- 
relation between the apparent increase in density and 
the amount of extracted SO* tYLus 0.05 K fioA _ rr) i« a**- 



shown 



Other Properties 



Among the tests that show wholly insignificant de- 
grees of correlation with the bleeding rate are the Paul 
floe test and the Merriman sugar test. Both tests show 
some correlation with the bleeding capacity, but the 
odds against chance correlation are too low to give the 

indications much significance. Results are reported in 
Table 8, below. 



-U3 



Table 8 



S 



Results of Statistical Analysis 



Test on Cement 



Paul 
Floe 
Test 



Merriman 
Sugar -Test 



Phenol - 
phthalein Clear 
End Point Point 



No. of cements 



23 



23 



23 



Coefficient of 
c orrelat ion 
with bleeding 
rate : * 



0.03 



- 0.10 



- 0.10 



Odds against chance 
correlation with 



bleeding rate: 



1 to 8 



3 to 5 



3 to 5 



Coefficient of 
correlation 
with bleeding 
capacity: 



0.30 



0.40 



- 0.40 



Odds against chance 
correlation with 
bleeding capacity: 



6 to 1 



16 to 1 15 to 1 



*The bleeding data used here are the same as in Fig. 13. 



-U- 



Summary 

Correlation of bleeding data with data on heat lib- 
eration leaves little doubt but that the bleeding is 
materially affected by chemical reactions that occur when 



cement 
ium al 



Reactions of tri 



the relatively great reactivity and heat -producing ca- 
pacity of this compound. Correlations between other 
properties and the bleeding are readily explained in 
terms of reaction of the aluminate. Apparently, the 
amount of the reaction-product is related to the amount 
of sulfate that goes readily into solution, and to a 
lesser degree to the amount of alkali in the cement, 
(which often supplies a considerable part of the soluble 
sulfate). The amount of the reaction product appears to 
increase with the amount of tricalcium aluminate in the 
cement when this is above 7<£. 



In general, the bleeding rate gives higher coeffi- 
cients of correlation with other properties of the ce- 
ment than does the bleeding capacity, but the coeffi- 
cients have the same signs. 

8. Miscellaneous Influences Affecting 

Rate and Bleeding Capacity 




The bleeding characteristics of a cement appear to 
be subject to as many influences as are other physi- 
cal properties such as plasticity, setting time, and 
strength. Some miscellaneous information on these va- 
rious influences has been assembled in this section. 

Rate of Clinker Coo 




Employing methods somewhat different from that of 
Bulletin 2, William Lerch (10) showed that slow cooling 
of clinker can reduce the amount of bleeding. More re- 
cently, the method of Bulletin 2 has been used with two 
pairs of cements ground from quick and slow cooled clink- 
ers. Some of the results are shown in Table 9. It is 



-45- 



evident that slew cooling can markedly reduce both 
bleeding rate and bleeding capacity, especially the lat- 
ter. 



Table 9 



Bleeding Rate and Capac 



As Affected b 



Rate of C linker Cooling 
Paired cements are from a given raw mix* 



- 



m 



Cement Clinker 



No. 



Cooling 



Sp. Surface, 
sci cm per gn 
(Turb. Air- 
Method) Permea- 
bility 
Method 



w 



Bleeding Bleeding 

Rat e Capac ity f 
Q x 10° AH f 



14901 
14903 



quick 

slow 



14910 

14912 



quick 
slow 



1640 
1630 

1610 
I64O 



3030 
3070 



O.59 184 
O.59 123 



0.59 
0.59 



184 
109 



0.092 
O.042 

O.065 
0.024 



Calcium Sulfate Retarder 



In Bulletin 2 an increase in the proportion of gyp- 
sum in the cement was shown to have a large effect In 
reducing both bleeding rate and bleeding capacity even 
though the concentration of sulfate in solution did not 
vary much and the specif ic surface (by turbidimeter) was 
maintained practically constant • Subsequent tests have 
supported these findings. A simple explanation can now 
be given as a result of determinations of specific sur- 
face by the air-permeability method. It has been shown 
elsewhere (13) that the ratio of specific surface as de- 
termined by the air-permeability method to that as de- 
termined by the A.S.T.M. turbidimeter increases consis- 
tently with increase in the S0 3 content of the cement. 
Probably the air -permeability method gives the best ac- 
count of the specific surface of the gypsum, since the 



-U6- 



turbidimeter method does not analyze below 7^ microns 
in particle-diameter.* Accordingly, when the specific 
surfaces, as determined by the turbidimeter, appear to 
be the same, the cements containing higher amounts of 
gypsum are probably actually finer and show lower val- 
ues of Q and AH 1 for that reason* 



gypsum can 



ir it is partially dehydrated during milling. Lerch (10) 
investigated the effects of such dehydration and found 
that the bleeding capacity was greatly reduced, provided 
the paste was mixed for just 2 minutes. Longer mixing, 
conforming to the method of Bulletin 2, brought the 
bleeding capacity back almost to the same order of mag- 
nitude as was obtained when there was no dehydration. 
Lerch did not study bleeding rate, but more recent tests 
made according to the method of Bulletin 2 indicate that 
the dehydration causes a pronounced reduction in the 
rate. The tests were made on ground clinker, with the 
addition of plain gypsum in one case, and the same gyp- 
sum, partially dehydrated, in the other. The results 
are shown in Table 10. 



In further work, a cement ground in the laboratory 
was compared with a commercial cement that was believed 
to contain dehydrated gypsum. In the laboratory grind, 
some of the same clinker was used and practically the 
same fineness was produced without dehydration of the 
gypsum. The results of bleeding tests, shown in Table 
11, are comparable to those in Table 10. 



"Other factors that affect the accuracy of the turbidi- 
meter analysis are the differences in density and in 
opacity between gypsum and clinker. 



-47- 



In other tests a limestone powder was mixed with 
plain gypsum in one case, and with dehydrated gypsum 
in another. The preparation containing the dehydrated 
gypsum gave much the lowest bleeding rates and bleeding 
capacities. Apparently, pronounced dehydration of the 
gypsum always tends to reduce the bleeding rate. Whethr 
er it reduces the bleeding capacity depends upon the 
mixing schedule, and perhaps upon other circumstances 
also. 



Table 10 



Bleeding Data for Ground Clinker 
No. SBR - 16065 at w « 0.65* 



Bleeding Bleeding 
Retarder Rate / Capacity, 
Q x 10 AH' 

Dehydrated gypsum 204 O.O85 

Plain gypsum 280 0,085 



Table 11 



Bleeding Data at w ^ 0,575 



rn^m 



Bleeding Bleeding 
Cement Rate Capacity, 
Q x 106 AH 



t 



Mill cement SBR - 

15667 believed to 

contain dehydrated 

gypsum 230 0.128 

Lab. grind from same 

clinker as cement 

SBR - 15667 298 0.131 



-1,8. 



Aeration of C em out 



might 



Leroh (10) found that alkali carbonates such as 

ent greatly reduced 



ceding capacity of a paste mixed only 
mixing increased the bleeding capacity 
xg it back to the original value wfc 
amounted to 0.9& or more. 



admixed 



Some cements tested according to the method of 
Bulletin 2 have shown large reductions in bleeding rate 
and bleeding capacity after storage for a few months 
in partly filled containers that were not air-tight. 
Sometimes only the one or the other property was af- 
fected* 



Salts 



In Bulletin 2 it was indicated that dissolved salts 
reduce the bleeding rate and may also reduce the bleed- 
ing capacity. Many tests have been made since then 
involving various cements and various alkali salts, but 
chiefly sodium chloride. The salts always reduced the 
rate but generally had little effect on the bleeding 
capacity. Special tests indicated that the effect on 
the rate was less if the gypsum in the cement was strong- 
ly dehydrated. Reductions in rate were commonly some- 
what less than those found in the tests reported in 
Bulletin 2, especially at the lower concentrations of 
salt. " A ret est of cement 14-502, which had supplied 
data for Bulletin 2, showed a very much diminished ef- 
fect, indicating some change during storage. A few of 
the data for other cements are given in Table 12. 



-49- 



Table 12 



Effect of Sodium Chloride in 1 Normal Solution 
— "" (a 5.6% solution) 



Cement 



NaCl 
in Mix 



Bleeding IJate, 
Q, x 10 




w Actual Calculated 



Water 






for same 
density and 
viscosity as 

pure water 


Bleeding 

Capacity, 

AH' 


SBR-15621 


none 


0.62 
0.62 


224 
177 


195 


0.082 
0.069 


SBR-15931 


none 


0.60 
0.60 


151 

103 


134 


0.089 

0.083 


SER-15668 


none 

5.6* 


0.63 
0.63 


270 

220 


243 


0.H8 
0.H8 


SBR-15667* 

* 


none 
5.6* 


0.55 
0.55 


167 
151 


166 


0.099 
0.105 



*Believed to have contained dehydrated gypsum 

• As was noted in Bulletin 2, calcium chloride pro- 
duces larger effects per chemical equivalent than the 
alkali salts. Some recent data on the effect of calcium 
chloride are shown in Table 13 . It is evident that the 
bleeding capacities are more strongly reduced than the 
initial bleeding rates. This, together with the fact 
that the calcium chloride also greatly reduces the bleed- 
ing time, supports the view expressed in Bulletin 2 that 
the influence of the calcium chloride may be largely the 
effect of its known acceleration of setting. It is of 
interest in this connection that the constant-rate pe- 
riods tended to be rather brief when the calcium chloride 
was present. Another effect of the calcium chloride was 
to make the cements more susceptible to channeling. 



50- 



Table 13 



Tests with and without Calcium 
Equal to 1% of the Weight of Cement 

2.1$ of the water 



60 mm high, at 23.5°C and 
cement SER -15668 which wa 



tested at w « 0.575. 



Bleeding Rate, 
Q, z 10 6 



p™™+ Bleeding Actual Calculated ,„,,,_ 

Cement a****.* mJ * Bleeding 

Addition Time, for same capacity 

min. density and 



No. 



viscosity as 
pure water 



Ah 1 



ip 



SBR-15365 none 71 163 O.O85 

CaClg 38 146 156 0.040 

0.049 
95 0.020 



S BR -15496 


none 
CaCl 2 


46 
26 


129 

89 


SBR-15621 


none 

CaCl 2 


71 
46 


139 
103 


SBR-15697 


none 
CaCl 2 


81 

58 


122 
94 


SBR-I5668 


none 
CaCl 2 


87 
71 


156 
122 


Alkali Hydroxides 







0.072 

110 0.039 



0.077 

100 . 047 



0.093 

131 . 068 



«M»aaWM»a4 



It was .previously reported (in Bulletin 2 t p. 51) 
that bleeding rate is not affected by sodium hydroxide in 
solution. This matter has been further investigated 
using two cements, both of which were low in alkali 
oxides (less than 0.35$). The initial solutions were 
1/8, 1/4, and 1 Normal in alkali hydroxide, and both NaOH 
and K0H were used, in different tests. After mixing the 



51- 



solutions with the cement, the hydroxide concentrations 
fell markedly because a high concentration of alkali 
hydroxide reacts with gypsum and precipitates hydrated 
lime, as shown by the following equations:* 

CaSO^ + 2NaOH m NaeSO* + Ca(0H)g 
CaSO^ + 2K0H « E2SO4 + Ca(OH) g 

When the weaker solutions (1/8 and 1/4 Normal) were used, 
the bleeding rates and bleeding capacities were not 
strongly affected; but there was, in some cases, an ap- 
preciable increase in rate and decrease in bleeding capa- 
city. The 1 Normal hydroxide reduced the bleeding rate, 
but no more than corresponded to the greater density and 
viscosity of the liquid. It materially reduced the bleed- 
ing capacities of both cements, but the effect was much 
greater for one cement than the other. As the cement most 
affected was high in C3A whereas the other was not, it 
seems likely that the reduction of the amount of retarder 
by reason of the reaction cited above may have caused the 
difference in results. Some of the bleeding data obtained 
with the cement on which 1 Normal hydroxide had the least 
effect on the bleeding capacity are shown in Table 14* 



The tendency of the 1 Normal hydroxide to reduce the 
bleeding capacity while producing no specific effect on 
the bleeding rate is analogous to results obtained with 
silica powder when the concentration of flocculant is 
increased (See Bulletin 2, p. 72). To study the possibil- 
ity that the hydroxide affects the state of flocculation, 
some auxiliary tests were made in small vials using 2 
grams of cement to 10 ml of water or solution. A -g- Normal 
solution of NaOH caused the sediment to occupy more than 
twice the volume that it had in tap water, and the super- 
natant liquor was brilliantly clear, whereas it was not 
when tap water was used. Other concentrations also gave 



analys is 



-52- 



Table 14 
Ef fect of Alkali Hydrox ides on Cement No. SBR-1S668 

m ■ ■ ■■ ' ■ ' " ■ ■ r ■ ■ ■ ■ ' >■■■■■ ~.i ffi i ■■ ■ ■■ ■ ■ ■ ■■■■■■■■■ rfV M 

23.5°C) 



Bleeding Rate, 

Q x 10 6 

Solution w Actual Calculated Bleeding 

for same Capacity, 
density and AH* 
viscosity as 
pure water* 

^^ ' — ■ — * • — — ■ **— ¥- n t-i — — ■ 



water 0.575 l6l 0.105 

1/8 N. NaOH " 164 0,105 

1/4 N.' NaOH " 152 0.096 

1 N. NaOH " 123 157 0.067 

1/8 N. KOH w 162 0.086 

1/4 N. KOH " 164 0.083 

water O.O63 260 0.160 



it 



1/8 N. NaOH " 283 0.169 

1/4 N. NaOH f * 277 0.153 

1 N. NaOH " 211 270 0.086 

1/8 N. KOH " 297 0.142 

1/4 N. KOH " 304 0.143 

1 N. KOH " 219 251 0.079 



*The correction is relatively small for the 1/8 and 1/4 
N solutions. Since it is a slight over-correction with 
respect to cement liquor, which is not pure water, it 
was not applied to these weaker solutions. 



-53- 




volumes of sediment greater than in tap water, but the 
\ Normal solution showed the greatest effect • In no 
case did the sediment "set" during the tests. It is be- 
lieved that the alkali hydroxides probably do affect the 
state of flocculation, and that the hydroxyl ion, which 
can be supplied by calcium hydroxide also, may perhaps 
act directly as the flocculant. The cement particles in 
a paste are normally positively charged in spite of the 
fact that the paste is flocculated* \ perhaps it is the 
negative hydroxyl ion that keeps the charge sufficiently 
low to permit of the flocculation. 

rature During Test 

Tests reported in Bulletin 2 showed that with in- 
crease in temperature a reduction in rate of bleeding 
began at about 25°C in spite of the decrease in viscosity 
of the water with temperature. At lower temperatures the 
rate varied in inverse proportion to the viscosity of the 
water as would be expected from Equation 2 if w^ remained 
constant. Because of the unexplained reduction in rate 
at the higher temperatures, most of the subsequent bleed- 
ing tests have been made at about 23.5 ♦, 1°C. However, 
because cements are often used at higher temperatures the 
cements of the ITS group have been tested at two different 
temperatures, namely, 23«5°C and 32°C (74°F and 90°F). 
At both temperatures enough tests on each cement were 
made at a water-cement ratio of 0.^66, by weight, to es- 
tablish the bleeding rates and bleeding capacities rather 
well. The data are shown in Table 15. It is evident that 
in most cases the bleeding rates increased with temper- 
ature, and that the cement which gave the data reported 
in Bulletin 2 was therefore not representative of cements 
in general. 



The ratios of the bleeding rates for 32°C to those 
:3.5°C would have been 1.20 had the change in rate 
change in temperature resulted onlv from the change 




■VBV^ 



his was sham by electroosmosis in tests that the writ- 
er made for the Riverside Cement Co. The reference is 
to pastes not over 7 hours old. 



54- 



Table 15 



Bleeding Data for Cement Pastes at 23»5»C, and 32*0 
( Water-cement ratio, by weight - 0*466) 



Cement 


Initial 
23.5°C 


Bleeding 
32°C 


Rete, Q 1 106 
32° values 
23. 5° values 


Ble« 
23. 5*C 


sding Capac 


ity. ^H' 


No. 
(LTS group) 


32* C 


32° values 
23.5" values 


11 - 616 
11T- 717 

12 - 625 


120 
62 

120 


134 

160 


1.12 

1.33 


0.080 
0.030 
O.O84 


O.O65 
0.081 


0.81 
0.96 


12T- 726 

13 - 621 

14 - 423 


127 
134 

75 


80 

159 

66 


O.63 
1.19 

0.88 


0.038 
0.062 
0.050 


O.046 

0.053 
0.042 


1.21 
0.85 
O.84 


15 - 526 

16 - 834 
16T- 835 


130 

139 

115 


170 
167 
131 


1.31 

1.20 

1.14 


0.079 

0.086 
0.069 


0.080 
0.083 
0.067 


1.01 
0.96 
0.97 


17 - 640 

18 -2044 
16T-3025 


86 
130 
105 


120 
115 


1.40 
1.10 


0.059 
0.078 

0.044 


0.067 
0.023 


1.13 

■ 

0.52 


21 - 223 
21T- 310 

22 - 618 


190 
120 

141 


217 
151 

176 


1.14 

1.26 

1.25 


0.120 
0.085 
0.053 


0.120 

0.079 
0.076 


1.00 

0.93 

1.43 


23 - 521 

24 - 631 

25 - 655 


I48 

71 
109 


173 

60 
130 


1.17 

O.84 

1.19 


0.091 

0.053 
0.086 


0.090 
O.042 
O.O85 


0.99 
0.79 
0.99 


31 - 628 
33 - 430 
33T- 511 


35 
54 
36 


46 

57 
53 


1.31 

1.06 
1.47 


0.015 
0.025 
0.015 


0.012 
0.023 
0.017 


0.80 
0.92 

1.13 


34 - 621 
U - 660 

42 - 630 


60 

76 

146 


71 
86 

162* 


1.18 

1.13 
1.11 


0.029 
0.061 

0.114 


0.029 
0.056 
0.112 


1.00 

0.92 

0.98 


43 - 722 
43A- 322 
51 - 532 


75 

89 

112 


93 
112 
I83 


1.24 
1.26 

I.63 


0.053 
0.053 
0.077 


O.O56 

0.050 
0.111 


1.06 

0.94 
1.44 



A "T" in the cement number indicates a so-called "treated" cement, that 
Is, an air-entraining cement made by addition of Vinaol resin* 



-55- 



in viscosity of the water. For the 20 cements in Table 
15 that were made without Vinsol resin and for which 
data are available at both temperatures (that is, not 
counting cement No. 18) the average ratio is 1.20 and the 
ratios for 14., or 70$, of these cements are within 10$ 
of the theoretical value. It appears that in most cases 
the theoretical rate -equation, with about the. same value 
of wi, remains applicable as the temperature is raised 
considerably above 23°C. However, some cements do not 
conform at all well, and this is probably attributable 
to changes either in the amount of early chemical reac- 
tion or in the state of flocculation. 



The ratios of the bleeding capacities at the two 
test temperatures average 0.99 for the same 20 cements, 
and 12, or 60$, of these ratios are within 10$ of 1.00. 
In general, therefore, there was little change in bleed- 
ing capacity with the change in temperature, but again 
there were some pronounced exceptions to the general rule. 



Vinsol Resin 



The air-entraining cements of the LTS group were all 
made by addition of Vinsol resin (about 0.04$) during the 
grinding. Cements made at the same plants from the same 
clinkers but without this addition provide a good basis 
for judging the effect of the resin. In Table 15, cements 
made from the same clinker are indicated by the same num- 
ber in advance of the hyphen, except for the T, which de- 
signates the air-entraining cements. Table 15 contains 
only a few of the bleeding data which are available for 
comparison. 



In Bulletin 2 a method is given for estimating the 
effect of air-entraining admixtures like Vinsol resin 
on the bleeding rate of a cement. According to this 
method the entrained air is treated as a weightless ag- 
gregate of negligible surface exposure. Results obtained 
with an air -entraining agent were shown in Bulletin 2 to 
agree rather well with the calculated values. Similar 



-56- 



calculations were made for the treated, or air-entrain- 
ing, cements of the LTS group, but the agreement was 
not always good. The principle of the calculation is 
believed to be valid, but evidently something more is 
somet imes involved - 



capacity by more than can be 



eding 



Ljuuiroeu. ior oy xne volume 
occupied by the air itself. Air contents of the pastas 
of Table 15 that were made with treated cements were 
found experimentally to range from 1.2% for cement 11T to 
13.2JS for cement 16T. 

9« Summary and List of Factors That Affect 

the Bleeding of Cement Paste 



1. Subsequent experimental and theoretical studies 
have supported, in general, the views on the mechanism 
of the bleeding of cement paste that were presented in 
P.C.A. Bulletin 2. However, because of minor revisions 
of theory, the equation for initial bleeding rate that 
was developed in Bulletin 2 now appears to represent cer- 
tain basic influences rather less precisely than was at 
first thought. Nevertheless, it provides a sufficiently 
accurate representation of experimental data. 

The later work has supported the previous explanation 
that Wi, the experimental constant of the rate equation, 
is required because a part of the water remains immobile 
with respect to the cement particles, i.e., does not flow 
past them during the bleeding. However, w^ is now con- 
sidered to be only an indirect measure of the quantity of 
such water. Also, most of this water is now believed to 
be present in "dead" spaces caused by the angularity of 
the particles, rather than in thick, solvated layers bound 
to the particles, as was previously considered probable. 
However, it is thought that thinner layers, of hydration 
products, probably do form around the cement particles 
when they first contact water, and are largely responsible 
for the differences between the vt^ values. Support for 
this latter view is found in the correlation of the data 



-57 



on bleeding with the initial heat liberation, and with 
compositional factors. 

2. The fact that a cement paste is flocculated is 
now given more emphasis than formerly, especially in 
analyzing the influences that affect the bleeding capa- 
city. New experimental studies have led to revision of 
the equation relating the bleeding capacity to the water 
content of the paste. The following equation, though 
purely empirical, appears to be generally satisfactory: 

AH' = Jc^c [w/c - (w/c) m ] e 

The symbols are defined in the nomenclature list at the 
end o f this r eport. The equation gives a straight line 
whenV^H'/c is plotted against w/c. 

3. The rather extensive bleeding data now available 
have been analyzed with respect to the influence of cement 
fineness, the relationship between bleeding rate and 
bleeding capacity, and the .duration of bleeding. The 
earlier indications regarding the strong influence of the 
cement fineness on the bleeding cteracteristics have been 
confirmed; reduction in water content and increase in ce- 
ment fineness are, in general, the most Important ways of 
reducing the bleeding of normal, untreated portland cement 



paste. Bleeding rates and bleeding capacities have been 
shown to be closely related but, when different cements 
are concerned, a difference in rate is by no means an 
infallible indication of a difference in bleeding capa- 
city. Duration of bleeding has been found to be remark- 



ably insensitive, in many cases, to the water content of 
the paste. 



4. Among miscellaneous inf. 
temperature of the paste has ] 



have now been made 



at 74 and 90°F. Most cement pastes show about the same 
bleeding capacity at both temperatures, and they change 
in bleeding rate to about the extent that corresponds to 



-58- 



the change in viscosity of the water. However, a con- 
siderable number of cements change their bleeding char- 
acteristics with change in temperature in a way which 
is unpredictable at present though probably related to 
the amount of initial reaction, and perhaps also to the 
state of flocculation. 



5. Tests of cements treated with Yinsol resin in- 
dicate that the effect of the entrained air is not al- 
ways fully predictable simply on the basis that the air 
is weightless aggregate. The entrained air generally re- 
duces the bleeding capacity of the paste by more than 
the volume occupied by the air. 

6. On the basis of the information now available, 
it is possible to list a considerable number of factors 
that affect the rate and amount of bleeding of a cement 
paste. Important among the test conditions are, of 
course, the water content and the temperature of the 
paste. Factors that relate to the composition, produc- 
tion, or storage of the cement are listed below. Al- 
though the magnitudes are so expressed that lowering of 
the bleeding rate and bleeding capacity is the effect 
that is indicated, it will be evident that some of the 
factors could be varied in magnitude in such manner as to 
increase bleeding. 



Can 



and Bleeding Capacities 



of Cement Pastes 



high soluble alkali in clinker 
high C3A in clinker 
slow cooling of clinker 
high cement fineness 
large addition of gypsum 

hot milling 

addition of an air-entraining agent 
addition of calcium chloride 
exposure of cement to the air 



-59- 



This is not to be regarded as a complete list of all 
factors that can influence the bleeding. It covers only 
those factors on which enough evidence has been obtained 
to warrant mention at this time* 



60. 



Nomenclature 



■ absolute volume of cement per unit volume of mix 
eg = maximum volume of cement per unit of settled volume 

(the maximum obtainable with a given cement) 
d c ■ density of cement, gm/cm 3 
df = density of water, gm/cm3 

g ■ the gravitational acceleration, cm/sec 2 , (about 980) 
AH' = bleeding capacity. It is the ratio of the total 

decrease in paste height to the initial paste height* 
k - the slope of the data line in a plot of V AH f /c 

versus w/c 
Q = bleeding rate, cm/sec. It is the initial rate of 

settlement of the top surface of the paste. 
R = the ratio of the actual bleeding time to the time 

in which the same settlement would have occurred 

had the initial rate prevailed throughout 
T s « a velocity calculated by applying Stokes' law, 

expressed in terms of a, to a given powder, 

irrespective of whether the particles are of uni- 



form size 
w ■ volume of 



i) 



w-o ■ minimum volume of water per unit of settled volume 

(w B + c B « 1) 

w^ = a term (used in equations for bleeding rate) which 

has a constant value for a given cement tested at 
a given temperature. It is a correction for water 
not involved in the flow but does not represent 
directly the quantity of such water. 
( w /c)m = minimum water -cement ratio (by absolute volume) for 

pastes of a given cement. It is determined by 

Equation 7« 
N = Number of pairs of data 

q ■ coefficient of viscosity of water, gn/cm sec, (poises) 
O » specific surface of cement (on volume basis) , cm 2 / cm 3 
- specific surface of cement (on volume basis) , cm 2 / cm 3 , 
as determined in accordance with the Wagner turbidi- 
meter method. A.S.T.M. Designation C115-42 



a 



w 



-61- 



References 



Xm Bastow, S. H. and Bcwden, F. P. f "Physical Properties 
of Surfaces. II. Viscous Flow of Liquid Films; 
Range of Action of Surface Forces," Proc. Roy. Soc. 
(London) (A) lgl, 220^233 (1935). 

2m Bulkley, R. f "Viscous Flow and Surface Films," Bur. 
Stds. J. Res. 6, 89-112 (1931). 



on Buzagh, A., "Colloid Systems," Translation by 
Otto B. Darbishire, London: The Technical Press 
Ltd., 1937. 



4. Carman, P. C, "Fluid Flow Through Granular Beds," 

Trans. Inst. Chem. Engrs. 1£, 150-l6l (1937). 

5. Carman, P. C, "The Determination of the Specific 

Surface of Powders I," J. Soc. Chem. Ind. 57, 225- 
234T (1938). 

6. Carman, P. C, "Determination of the Specific Surface 

of Powders II," J. Soc. Chem. Ind. j>8, 1-7T (1939). 

7. Fair, G. M. and Hatch, L. P., "Fundamental Factors 

Governing the Streamline Flow of Water Through 
Sand," J. Am. Water Works Assn. 2£, I55I-I565, 

(1933). 



and 



Liquid Phase during the Early Stages of Hydration," 
Unpublished Report on the Long Time Study of Cement 
Performance in Concrete, Research Laboratory, Port- 
land Cement Association. October. 19 A3. 



9. Kozeny, Josef, "Capillary Conduction of Water in 
Soils," (In German) Sitzungsberichte, Akad. Wiss. t 
Wien, 136 > 2a, 271-306 (1927); ibid., "The Permea- 
bility of Soils," (In German) , Kulturt-echniker, 



2i, 478-486 (1932) 







-63- 



10, Lerch, William, "The Initial Reactions of Ifydration 

of Portland Cement and Their Effect on the Consis- 
tency or Water Requirement of the Cement Paste, 
Unpublished report of the Research Laboratory, 
Portland Cement Association, June, 1941 • 

11. Lerch, William, n The Influence of Gypsum on the 



Initial Reactions between Portland Cement and later, " 



Unpublished report of the Research Laboratory, 
Portland Cement Association, May, 1942. 

12. Peters, C. C # and Van Voorhis, M. A.,, Statistical 

Procedures and Their Mathematical Bases, New York: 
McGraw-Hill Book Co., Inc., 1940, pp. 91-99, 118- 
123, 152-155. 



13. Steinour, Harold H., "Experience with the Air-Permea- 
bility Apparatus for Measuring Surface Area of 



Cement," Unpublished report of the Research Labora- 



tory, Portland Cement Association, April, 1942. 

Steinour, Harold H. , "Rate of Sedimentation: Non 
Flocculated Suspensions of Uniform Spheres," Ind 
Eng. Chem. ^6, 618-624 (19W). 

Steinour, Harold H. , "Rate of Sedimentation: Suspen 
sioEs of Uniform-Size Angular Particles," Ind. Eng 
Chem. 26, 840-8A7 (19U). 



16. Steinour, Harold H. , "Rate of Sedimentation: Con- 
centrated Flocculated Suspensions of Various Pow- 
ders," Ind. Eng. Chem. ^6, 901-907 (19U). 



-64- 



Appendix A 



Chemical Constitution 



The chemical constitutions of the clinkers fr 



tut 



which were produced the cements having the group desig- 
nation SBR* are given in Table 16. The cements con- 



tained amounts of SO3 which varied only from 1.73 to 
2.11#, except for 2. 4856 in Lot SBR-15365, and 1.59% in 
Lot SBR-15o22# The specific surfaces of the cements are 



reported, along with the bleeding data, in Appendix B 
which gives the serial numbers of the cements produced 
from each clinker. 



Since most of the cements of the LTS group were 
ground from different clinkers, the chemical constitu- 
tions of the cements (rather than of the clinkers) are 
given for this group, in Table 17. Cements having the 
same identification except for a T were produced at the 
same plants, and from clinkers of approximately the same 
constitutions. The T designates air-entraining cements 
produced by addition of Vinsol resin. 



''"With respect to group-designations see page 2 



-65- 



Table 16 



Chemical Constitution of Clinkers from which 



-■■— i 



the Cements of the SBR Group were Produced 

The compounds were computed on the potential compound 
basis without correction of the CaO, Si02, AI2O3, and 
FegOa analyses, except for subtraction of the free 
lime from the total CaO. The letters C, S, A, and F 
are used to represent the oxides in the formulas for 
the four major compounds. 




Clinker C3S CsS C3A C4AF MgO **T" S0 3 ^f 11 L ? 1 
Ref.No. * % % % % CaO % Oxides, Ign. , 




1 46.5 27.6 14.3 6.8 2.6 1.0 0.4 0.3 0.2 

2 62.9 12.6 11.1 7.7 H 2.7 0.2 0.7 0.9 

3 47.6 3<3.9 4.8 13.2 1.4 0.7 0.0 0.2 1.1 

4 37.1 52.2 2.3 5.7 1.8 0.2 0.2 0.3 0.3 

5 48.4 28.7 10.1 7.6 3.4 0.0 0.2 1.6 0.4 



*Taken from cement analyses. 



-66- 



Table 17 

Ch emical Constitution of Cements 

of the LTS Group 



The compounds were computed on the potential compound basis without 
correction 'of the CaO, SiO g , Al , and Fe g 3 analyses, except for 
subtraction of the free lime from\he total CaO, The letters C, S, 
A, and F are used to represent the oxides in the formulas for the 
four major compounds. 

















Free 


Alkali 


Loss 


Cement 


C 1 S 


C £ S 


C A 


C 4 aF 


MgO 


CaSO^ 


CaO 


Oxides 


on 


No. 


% 


^ 


1 


* 


# 


* 

2.7 


0.4 


0.7 


Ign. t 


LTS -11 


51.0 


21.0 


12.1 


7.2 


3.6 


1.0 


LTS-11T 


5l.o 


21.0 


12.2 


7.1 


3.7 


2.6 


0.5 


0.7 


1.2 


LTS -12 


45.0 


28.0 


12.6 


7.3 


3.1 


2.7 


0.1 


0.7 


0.6 


LTS-12T 


46.0 


27.0 


12.5 


7.1 


3.1 


2.7 


0.1 


0.7 


0.7 


LTS -13 


50.0 


26.0 


10.1 


6.5 


l.l 


2.8 


1.6 


0.2 


1.8 


LTS-14 


A2. 5 


32.0 


8.2 


9.2 


2.5 


2.9 


0.2 


1.4 


0.9 


LTS -15 


64.5 


10.0 • 


12.1 


7.5 


0.8 


3-2 


0.4 


0.3- 


0.9 


LTS -16 


53.5 


21.0 


7.5 


10.7 


2.0 


2.9 


0.7 


0.7 


1.0 


LT5-16T 


52.5 


22.0 


7.9 


10.4 


2.0 


3.1 


0.8 


0.7 


1.2 


LTS-17 


52.0 


23.0 


10.4 


9.3 


1.1 


2.9 


0.4 


0.5 


0.9 


LTS-18 


44.5 


28.0 


13.2 


6.8 


2.1 


3.1 


0.3 


0.3 


1.0 


LTS-18T 


44.0 


28.5 


13.2 


6.7 


2.1 


2.9 


0.4 


0.3 


0.9 


LTS-21 


40.0 


41.0 


6.4 


9.7 


1.3 


2.1 


0.6 


0.6 


0.7 


LT5-21T 


38.0 


40.0 


6.6 


9.7 


1.3 


2.2 


0.6 


0.6 


0.6 


LTS-22 


41.5 


33.5 


6.6 


11.7 


3.2 


2.4 


0.1 


0.6 


0.6 


LTS-23 


51.0 


24.0 


3.7 


16.6 


0.9 


2.6 


0.4 


0.7 


0.6 


LTS -2 4 


41.0 


29.0 


5.4 


14.8 


3.1 


3.0 


0.9 


1.4 


1.4 


LTS -2 5 


34.0 


39.0 


4.7 


14.9 


2.2 


3.2 


0.2 


0.8 


0.6 


LTS -31 


56.0 


17.0 


10.8 


6.4 


3.3 


3.8 


1.4 


0.5 


1.5 


LTS-33 


60.0 


13.0 


10.4 


7.7 


1.4 


3.9 


1.8 


0.7 


1.5 


LTS-33T 


57.0 


16.0 


10.4 


7.3 


1.5 


3.7 


1.8 


0.7 


1.4 


LTS-34 


64.O 


10.5 


5.7 


10.1 


2.5 


2.9 


2.3 


0.6 


1.5 


LTS -41 


20.0 


51.0 


4.5 


15.2 


3.0 


3.4 


0.4 


1.3 


1.3 


LTS -42 


•27.0 


55.0 


3.5 


8.2 


1.8 


2.6 


0.2 


0.4 


0.9 


LTS -43 


25.0 


48.0 


6.2 


13.8 


1.6 


3.6 


0.1 


1.1 


0.7 


LTS -43 A 


29.0 


52.0 


5.3 


9.3 


1.1 


3.2 


0.4 


0o3 


0.8 


LTS-51 


41.0- 

■ 


39.0 


3.7 


10.0 


1.7 


2.4 


0.5 


0.3 


0.8 



-67- 



Appendix B 

Bleeding Data for Pastes Made with 
Cements of the SBR Grou 

In this appendix are presented all the data from an 
extensive series of bleeding tests made when the cements 
of the SBR group were first received. The cements have 
been studied systematically with respect to many proper- 
ties; the bleeding data reported here were obtained not 
only for the independent study of bleeding but to make 
comparisons with the other properties of the cements. 
The bleeding characteristics of the cements of the LTS 
group were determined with similar objects in mind. How- 
ever, they are not given here since Tables 3, 6, and 15 
of the main report are considered adequate for present 
purposes. 



The cements of the SBR group for which data are pre- 
sented here were ground at the plants. However, when ce- 
ments of four finenesses are represented, some were ob- 
tained by mixing. The data are in Table 18. The total 
bleeding times are shown only for those tests in which 
the times were determined to within 5 minutes. 



69 



Table 18 



Bleeding Data for Pastes made 

with Cements of the SBR Group 









Cement 




He ight 


Bleeding 


Bleeding 


Bleeding 






Specific 


Content 


Paste 


of 


Rate 


Capacity, 


Time , 


Cement 


Clinker 


Surface 


of 


Temp. , 


Paste, 


Q 1 10° 


AH' 


min. 


No. 


No. 


cm 2 /gm* 


Paste 
c 


°C 


mm 








SBR-15364 


1 


1040 


0.543 


24.0 


44.0 


92.5 


0.037 


45 








0.522 


23.0 


46.5 


118.0 


0.045 


40 








0.503 


24.5 


46.O 


158.0 


0.06 A 


46 








0.483 


24.0 


47.0 


208.0 


0.080 


45 








0.483 


23.3 


47.0 


249.0 


0.085 


40 








0.457 


23.4 


47-0 


263.0 


0.104 


45 


SBR-15365 


1 


16b5 


0.482 


23.7 


58.0 


59.2 


0.025 


55 








0.432 


24.4 


55.5 


103.0 


0.045 


55 








0.382 


22.8 


55.0 


189.0 


0,104 


76 








0.330 


22.4 


55.0 


312.0 


0.173 


71 


SBR-1 5366 


1 


2280 


0.480 


25.0 


58.0 


27.9 


0.007 


36 








0.430 


25.0 


49.0 


46.6 


0.016 


38 








0.380 


24.5 


57.0 


82.2 


0.035 


50 








0.350 


24.5 


44.5 


110.0 


0.055 


52 








0.280 

• 


23.0 


59.0 


262.0 


0.145 


70 


SBR-1 5495 


2 


1470 


0.510 


24.5 


33.5 


54.3 


0.028 


40 








O.486 


24.5 


33-5 


85.5 


0.042 


45 








0.461 


23.5 


45.5 


116.0 


0.059 


53 








0.436 


23.0 


34.5 


149.0 


0.076 


44 


SBR-1 5V?6 


2 


1740 


O.484 


23.8 


33.0 


49.0 


0.018 


33 








0.434 


23.2 


33.5 


84.0 


0.037 


37 








0.383 


23.0 


34.0 


I63.0 


0.083 


46 








0.335 


23.0 


34.0 


268.0 


0.146 


45 


SBR-1 5497 


2 


2500 


0.432 


25.0 


33.0 


23.8 


0.006 


20 








0.382 


24.5 


33.0 


54.8 


0.018 


25 








O.332 


24.0 


34.0 


106.0 


0.044 


32 








0.282 


23.5 


35.0 


173.0 


0.090 


40 


SHR*15620 


3 


1375 


0.530 


23.0 


33.0 


60.0 


0.032 


50 








0.505 


23.2 


33.0 


86.0 


0.054 


55 








O.48O 


23.0 


33.0 


129.0 


0.666 


48 








0.455 


22.5 


33.5 


159.0 


0.083 


54 


SBR-15621 


3 


1820 


0.499 


24.0 


32.0 


53.7 


0.033 


50 








0.458 


22.5 


45.0 


81,3 


0.051 


65 








0.419 


24*0 


46.0 


112.0 


0.078 


70 








0.378 


22.5 


46.5 


177.0 


0.118 


74 








0,329 


21.5 


46.0 


298,0 


0.194 


70 


SBR-1 5622 


3 


2200 


0.458 


23.8 


41.5 


43.4 


0.021 


55 








0.438 


24.5 


32.5 


55.7 


0.029 


44 








0.399 


23.0 


32.5 


80.7 


0.053 


55 








0.372 


24.2 


33.0 


116.0 


0.075 


55 



Specific surfaces were determined by the A«S,T.Me 
corrected for surface contributed by>325-«eflh 
significant only for the co*r»er cement e* 



turbidimeter method but were 
, The correction was 



-70- 



Table 18 (Concluded) 









Cement 




Height 


Bleeding 


Bleeding 


Bleeding 






Specific 


Content 


Paste 


of 


Rate 


Capacity, 


Time, 


Cement 


Clinker 


Surface 

A 


of 


Temp. , 


Paste, 


Q x 10 6 


AH' 


mln. 


No. 


No. 


£ / 
Cffi /gED 


paate 
c 


•c 


mm 








3BR-15622 


3 


2200 


0.351 


23.8 


33.0 


130.0 


0.087 


55 








O.322 


22.5 


47.5 


206.0 


0.139 










0.298 


23.8 


35.0 


289.0 


0.166 


55 


SBR-15&67 

ft 


4 


1250 


0.500 


24.0 


33.0 


109.5 


0.069 


9 « 


* 






0.475 


22.8 


33.0 


128.0 


0.083 


-- 








0.450 


22.5 


34.5 


177.0 


0.094 


■ 








O.425 


23.5 


34-0 


230.0 


0.128 


— * 


SBB-15762 


4 


1500 


0.490 


22.8 


32.5 


89.0 


0.056 


50 








0.460 


23.5 


34.0 


144-0 


0.080 


60 








0.430 


22.5 


33.5 


168.0 


0.102 


60 








0.400 


23.0 


34.0 


256.0- 


0.145 


60 


SBR-15763 


4 


1800 


0.450 


23.3 


32.5 


109.0 


0.072 


57 








0.425 


23.8 


33-5 


144.0 


0.094 


64 








O.4OO 


23.3 


34.0 


180.0 


0.113 


60 








0.375 


22.8 


34.2 


231.0 


0.117 


60 


SBR-15669 


4 


2290 


0.426 


24.0 


42.0 


83.3 


O.O56 


KM 








0.397 


24.2 


32.5 


97.8 


0.069 


— 








0.357 


23.0 


44.5 


153.0 


0.126 


— 








0.327 


23.6 


34.5 


197.0 


O.I64 


~ 


SBR-15696 


5 


1305 


O.48O 


24.0 


33.5 


108.0 


0.055 


45 








0.460 


24.5 


34.0 


154.0 


0.076 


50 








0.440 


23.5 


35.5 


- 189.0 


0.097 


45 








0.420 


24.3 


34.5 


229.0 


0.113 


45 








0.400 


23.0 


33.2 


310.0 


O.138 


46 


SBR-15760 


5 


1500 


0.450 


23.7 


34.0 


104.0 


0.056 


53 








0.400 


23.6 


34.5 


180.0 


0.118 


55 








0.375 


23.7 


35.0 


265.0 


0.151 


55 








0.350 


22.8 


35.5 


350.0 


0.187 


55 


SBR-15761 


5 


1800 


0.450 


24.6 


33.2 


67.0 


0.042 


58 








0.400 


23.8 


34.0 


107.0 


0.075 


60 








0.400 


23.1 


a. 3 


116.0 


0.090* 


75 








0.350 


23.4 


35.5 


202.0 


0.147 


65 








O.325 


23.4 


35.2 


251.0 


0.180 


— 


SBR -15698 


5 


2265 


O.45O 


23.8 


33.0 


34.8 


0.021 


50 








0.400 


23.5 


33.0 


66.0 


0.066** 










0.350 


23.3 


35.0 


118.0 


0.105 


72 








0.300 


23.7 


35.0 


205.0 


0.178 


80 



♦Result of two teats which showed 0.089 and 0.091 
**A check test showed 0.060. 



-71- 



Appendix C 

Additional Study of Equations 2 and 5 

The corrections for immobile liquid that are made 
in Equations 2 and 5 differ in one respect that is not 
discussed in the main text. This is, in deriving Equa- 
tion 5 from Equation 3 the quantity that was assumed to 

w i 

represent immobile liquid, namely c, was sub- 

1 " W i 

tracted from w and was added to c , whereas in the deriva- 
tion of Equation 2 only the w was corrected for immobile 
water. It follows that the two equations would differ in 
form even if equivalent assumptions were made as to the 
quantity of immobile water. It is the purpose here to 
show why the new approach to the problem requires the 
addition of the immobile water to c, and to show that a 
corresponding modification of Equation 2 would not con- 
flict with the method of derivation originally used in 
Bulletin 2. 



Equation 3* from which Equation 5 w &s derived, has 
been shown elsewhere (15) to be the same as 



w 3 

Q = 0.123V . — (9) 



where V is the velocity obtained by applying Stokes* 
law expressed in terms of o (rather than particle- 
radius), irrespective of whether the particles are of 
uniform size or not. It thus embodies the effects of all 
dimensional quantities. The rest of the right-hand mem- 
ber of Equation 9 shows how V is modified by the rela- 
tive proportions of settling Dodies and mobile liquid. 
Since V is determined by sedimentation analysis, it is 



-73- 



correct whether or not the solid particles carry liquid 
with them*. But liquid that accompanies the particles 

i+. hft firms irlfirftd a r>art of the mobile liauid. It 



ann 
unc 



volumes 



ettling and nonsettling constituents, both c and 
e corrected for any immobile liquid, as was done 



derivation of Equation $• 

In deriving Equation 14 of Bulletin 2, which is the 
same as Equation 2 of this report , the c was not cor- 
rected because it entered the equation simply as the 
factor by which the specific surface was multiplied in 
order to obtain the surface exposure per unit volume of 
paste. Indeed, from the viewpoint that was adopted, the 
c should not have been corrected, but the specific sur- 
face should, theoretically, have been corrected instead. 
Actually, in Bulletin 2 the specific surface was not cor- 
rected but was used as a first approximation to. the cor- 



rect surface, a procedure which appeared to be justified 
by the results. However, to make the rate equation con- 
form strictly to current theory the correction is needed, 
and it can be made as explained below. 



The studies that have now been made indicate that 
angular particles can be treated as spheres of solid 
plus immobile liquid. It is not supposed that the par- 
ticles with their accompanying liquid actually have true 
spherical shapes, but they act very much as though they 
did. Now, the specific surface o which is used in 
Equation 2 is calculated from Stokes 'law as though the 
particles were spheres, but the difference in density of 
the spheres and the liquid is assumed to be (d - dj. 



This is true for inert particles. If there is chemical 
reaction, the use of a different liquid in the sedi- 
mentation-analysis may make an appreciable difference. 
The present treatment neglects any such difference but 
does not differ in this respect from the treatment in 
Bulletin 2. 



-74- 



This makes no allowance for the fact that a "sphere" 

w. 

is not all cement but contains ^ — - — parts of liquid 

1 " W i 

to one part of solid, (In this development the immobile 
liquid is calculated in the same way as it was for Equa- 
tion 5,) The correct density difference for the spheres 
is 



w 

d + = — d 

c 1 - w. u f 

l 



- d 



1 + 



w i ~r 



1 - w 



or (1 - w^Md - d f ). Use of this density difference 

instead of (d - d_) in Stokes 1 law gives a V 1 - w. 

c f ° w v i 

for the specific surface of the spheres, if a is cal- 

w 

culated the same as for Bulletin 2, But o Vl - w 4 is 

W 1 

not the surface exposure per unit volume of cement, but 
per unit volume of spheres. Per unit volume of cement 



/ / w i \ a w V 1 - W4 

the surface is O yi - w, (1 + - — - — or -2- = 

W IV 1 - W . / "I - w 





1 - w. 

V * x ' 1 

The square of this, - , should therefore replace 

Q z in Equation 2. The result is then the same as if O 2 
w ^ w. w 

c i 

were left unchanged and - ■ . which is c + •= — — c . 

^^ 1 - w , ' 1 - w . * 

i i 

were used in place of c as was done in developing Equa- 
tion 5» To conform to the assumption that the immobile 

W i W i 

water is ■= c, - c instead of simply w. should 

1-w'l-w ^ * i 

be subtracted from each w of Equation 2, With these 
changes Equation 2 becomes the same as Equation 5 except 
for the value used for the empirical proportionality 
constant and the symbol used for the specific surface, 
(The a is assumed to be accurate whereas in obtaining 
(J the portion of the sample in which the particles 

are less than 7z microns in diameter is not analyzed,) 



-75- 



Some difference between Equations 2 and 5 might per- 
haps be expected to result from the fact that in Bul- 
letin 2 the buoyancy of the falling particles was treated 
as equal to the weight of the displaced water whereas in 
the derivation of Equation 5 it was shown to be greater 
and was therefore differently evaluated. However, the 
buoyancy was not actually needed in the development ac- 
cording to Bulletin 2. Over unit base area the weight 
of cement in excess of the weight of an equal volume of 
water was called the weight in excess of the buoyancy. 



but 
sure 



hydraulic 



-76- 



Appendix D 

Mathematical Developments with Respect to Bleeding 

cities 




Derivation of Equation 6 of Section 2 

Equation 6 expresses mathematically the hypothesis 
presented in Bulletin 2 (p, 65) that a cube of paste will 
settle until the height is the same as that of a cube 
having a volume equal to the limiting, or base, volume 
of the cement in the paste. This can be shown as fol- 
lows: 



Let the height of the cube of paste be 1. After 

settlement, the height will be 1 - AH 1 . The initial 

volume is 1, the base volume is (1 - AH 1 ) 3 , and the 

volume of cement is c. Let the volume of cement per unit 

c 

of base volume be represented by eg Then eg ~ (l - ^ 

or c o C3 (1 - AH*) 3 , which is Equation 6, 

The Assumed Linear Relationship between V A H' and w, for 

Coarse Cements 




The fact that for coarse cements the linear rela- 
tionship between V AH f and w holds at practically all 
possible concentrations suggests that the relationship 
represents approximately tho fundamental law governing 
the bleeding capacities of such cements. Consequently, 
a study of this relationship is presented here. 

The equation that represents the relationship is 
readily derived from the fact that the straight line pass- 
es through the points (1,1) and (wg t o), where wg is the 
volume of water in a unit of base volume.* Thus, the 



'he base volume that is here considered is toe minimum 
volume indicated by the assumed linear relationship 



between VA H T and w. It differs somewhat from 




urn 



-77- 



slope is — - ) and the equation is 

1 " W B 

• VSIF- b 



r^ 



W B 



Since w « 1 - c and w- * 1 - c fi , the equation may 
also be written 

t C R " C C 

VaiF- -5 — - i - 



C B C B 



The next point that will be examined is the rela- 
tionship between the paste concentration and the set- 
tled volume per unit volume of cement. This latter 

1 - AH' 
quantity is • But, from the above equation 



AH' = (1 £- ) e = 1 - £2_ + -<* 

C B c B c B 




Hence. 



1 - AH' 2 




C B C B 



(10) 



Thus the settled volume per unit, of cement decreases 
linearly with c, and since c » 1 - w it increases line- 
arly with w. 



A somewhat different insight into the significance 
of Equation 10 may be developed by rearranging the equa- 
tion as follows: 



° °B C B C B 

1 - AH' - — ♦ — (l - -2- ) 

C B C B C B 



(1 - AH') - 



°B 



o c 



1 - — =- "B (11) 



C B 



78- 



. To aee the significance of Equation 11, note that 
since l/c is the base volume per unit volume of cement, 
c/c R is tne base volume per unit volume of initial paste, 
1 - c/c R is the water in excess of "base" water in a unit 
of initial paste, and (1 - AH') - c /c R is the amount 
of this excess water that remains in the settled paste. 
Accordingly, Equation 11 shows that the proportion of the 
"excess" water that is retained in the paste in the same 
as the proportion of the initial volume that can be con- 
sidered "base" volume. See the closing part of section 
3 for other ways of expressing this. 



-79- 



Appendix E 



Measurement 



and the experi 



Bulletin 



reliable procedure for determining the bleeding char- 
acteristics of pastes can be outlined. If the test is 
made as described, at a cement concentration at which the 
paste can be poured easily but has sufficient coterence 
to avoid "channeling" during ' 
bleeding capacity will nearly 
2 and 7 which are re-oroduced 




bleeding rate and 
always conform to Equations 



ft 



0.2g(d c - d f ) (w - w i ) 3 
cr w 2 n * c 



(2) 



AH* « k 2 c [w/c .(w/c)J 2 (7) 

111 



The experimental constant w^ of Equation 2 can be 
determined frcm one bleeding test, but tests at two dif- 
ferent concentrations are required to determine the two 
experimental constants, k and (w/c) m of Equation 7. 
Tests at three or mare concentrations covering a lange of 
o l or mare in the value of c are desirable, in order to 
avoid long extrapolations in estimating results far other 
concentrations, and to test the accuracy of the equa- 
tions. (See Table 18 for suitable concentrations.) The 
equations will generally represent bleeding data with 
about the precision with which the data can be dupli- 
cated. However, neither equation should be regarded as 
strictly precise. Equation 2 is only partly theoretical 
and embodies practical approximations discussed in sec- 
tion 2; Equation 7 is wholly empirical. 



Lack of agreement with the equations will be indi- 
cated if data for several different concentrations cannot 
be represented adequately by straight lines on plots of 

(Qc) 1 '^ versus w, and A/ A H'/c versus w/c. Indeed, 



in the former case,. the slope of the line must also agree 
with a calculated value. (See later. ) Small discrepancies 



-81- 




• ;^# 



are to be expected, but when materially different re- 
sults are obtained the cement may be one that requires 
longer mixing in order to give "normal" results« If the 
paste tends to set quickly, normal results might 
haps be obtained at a lower paste height • If discrep- 
ancies of important magnitude persist after careful check- 
ing, it is obviously best to use the experimentally ob- 
tained curves in making predictions. However, if the 
testing is properly done, very few cases of this kin d 

are likelv to be found. 



The method of test follows: 



Suitable Equipment: 




Spatula 

50 or 100 ml 

250 ml class 



Thermometer, to 100 P C, graduated to ±" 
Balance, reading to 0.5 gm or better 
Electrical kitchen mixer, with one-quart 
mixing bowl. 

Lid for the one-quart mixing bowl, made so that it 
can be placed during mixing* It should have a slot 
(with cover strip) to admit the shafts of the stir- 
rers, and a hole through which a thermometer can 



ins ert ed 



during 



dimensions are A-in. diamet 



cm) and 3-in. height. The jar may be either glass 
or metal, but the walls must be rigid and the bot- 
tom flat. 

Gage for measuring the initial height of the paste. 
A straight length of wire used as a probe can after- 
ward be read against an ordinary scale to an accu- 
racy of about 0.5 mm. A gage giving greater pre- 
cision can be constructed from a vernier caliper. 




The caliper is mounted vertically (with jaws at top) 
on a T -frame that can be laid on top of the test 
jar. The main scale of the caliper is made immov- 
able with respect to the T-frame, A guide strip 
which moves with the movable jaw is tipped with 
a small flat metal plate which is brought into 
contact with the surface of the paste. With this 
device the level of the paste can be read to 0,1 



For a 40 mm depth of paste the precision is 

1/4J&. 

Float for use on the paste, A disk of bakelite or 
lucite 20 to 25 ma in diameter and about 2 ran thick 
can be used for the main body of the float, A 
thread of glass about 3-in, long is mounted ver- 
tically at the center of this with a bit of wai 
(Alternatively, a fine wire can be used, but the 
weight of the float divided by the volume of the 
disk must be less than the density of the paste,) 

25-power microscope (with cross-hair in eyepiece) 

mounted horizontally and equipped for vertical 
movement over a 20-to 30 ^m range with provisi 



for reading elevation to 0,001 mm. Such an instru- 




ment has ample accuracy and is obtainable, com- 
pletely equipped, frcm precis ion- instrument cam- 

, However, bleeding rates can be obtained 
to an accuracy of about 1% by using 5x magnifica- 
tion and reading to 0,005 ohu 

A stop-watch, or laboratory timer. An ordinary 
watch with a second hand can be used but is less 
convenient. 



Aght source against which to view the end or tne 
float-stem with the microscope. This may be a 10- 
watt lamp with a piece of paper for a diffusing 
screen. Avoid high-wattage lamps which heat the 



unn 



-83- 



300-ml Erlenmeyer flask, if density of cement in 
water is to be determined (see later discussion). 



Preparation of Paste 



A height of paste of 40 to 50 ™ appears to be most 
satisfactory.* In a jar 10 cm in dia # the corresponding 
volume of paste will be about 300 to 4,00 ml. Assuming 
350 ml, the amounts of water and cement needed are 35$* 
ml and 350cd« erams. respectively. 



It is desirable to run all tests at a constant tem- 
perature. If the laboratory temperature is not con- 
trolled, tests can still be run at a constant temperature 
by talcing suitable steps to minimize heat transfer. Most 
of the tests at the P.O. A. Research Laboratory have been 
made at 23-24°C. The mix water should be at a lower tem- 
perature, determined by experience, so that after mixing 
the paste-temperature will be within t 1°C of the chosen 
test temperature. 

Place the water in the one-quart mixing bowl and add 
the cement while the stirrers are turning at low speed. 
Use the spatula to bring into the mix any cement that 
clings to the sides of the bowl either above or below the 
fluid line. Put the lid in place and run the mixer at 
full speed for 2 minutes. Allow the paste to stand for 
3 minutes and then mix at full speed for 2 more minutes. 



*Use of such heights will sometimes result in total 
bleeding-times that exceed the apparent "dormant" 
period, but experience has shown that the estimated 
dormant period can be considerably exceeded without 
much affecting AH'. This is reasonable because 
only the density of the sediment last laid down is 
affected. Paste heights of 20 mm are too low and 



give low results, both for Q and AH 1 . A 30 
height is much better, but the results still tend 
to be a little low, especially the value of AH'. 



-84- 



During mixing, keep the bowl turning slowly. Take the 

temperature of the paste shortly before shutting off the 
mixer. 



After the mixer is shut off, keep the paste stirred 
with the spatula, or a spoon, and transfer it as quickly 
as possible to the test-jar. Start the timer and Immed- 
iately measure the height of the paste. If a plumb wire 
is used, lower it vertically into the paste and rotate 
it a bit after it touches bottom. Then lay it aside and 
measure the coated length later with a millimeter scale. 

Measurement o f Bleeding 

Place the jar of paste in front of the microscope 
and pour water from the small graduate* down the side of 
the jar till it forms a layer over the surface of the 
paste. Submerge the float-disk in the water at the cen- 
ter and let it down upon the surface of the paste. If 
the disk does not wet readily, a pencil can be used to 
force it under the surface of the water. After releasing 
the float, jar the vessel to seat the float properly. 

Place the light-source back of the test -jar, focus 
the microscope on the tip of the float-stem, and set the 
cross-hair for a scale reading. Take readings at one- or 
twonninute intervals during the period of constant bleed- 
ing rate. Thereafter, readings at five-minute intervals 
are sufficient to determine the shape of the bleeding 
curve. When only the initial bleeding rate and the 
bleeding capacity are of interest, readings need betaken 
only at 10 or 15-minute intervals after the initial rate 
has been established. 



*A bottle of water, mounted at slight elevation and dis- 
charging from a bottom outlet through a length of 
rubber tubing, is a convenience if many tests are 
made. 



85- 



After each scale reading, the difference from the 
immediately preceding reading and from the initial one 
can be figured and entered in the record. The difference 
from the initial reading is a convenient value to plot. 
Differences between successive readings show how constant 
the rate is and enable the end of the initial constant- 
rate period to be determined . 

The Bleeding Data 

The initial bleeding rate that remains constant for 
some 10, 20, or more minutes at the start is the Q of 
Equation 2. It is best determined from a plot of the 
change in height of paste versus time* Occasionally, 
the first one or two points are a little out of line, in 
which case they should be disregarded. The straight line 
whose slope gives the value of Q is extended to the zero 
of the time-scale, and the total settlement is figured 
from this zero position. This total settlement divided 
by the initial height of the paste gives the bleeding 
capacity AH 1 , 

The w i of Equation 2 is determined from a plot of 
(Qcp-/^ y ersus w# The equation indicates that the data- 
points should fall along a straight line which cuts the 




s of w at w^. The theoretical slope of the line is 

... . - 0.2g(d c - df) . The data-lines are 

the cube-root of Jl _. 

tf z n 

w ■ 

always drawn to the theoretical slope except when there 
are obvious discrepancies of significant magnitude. 

The constant k of Equation 7 is the slope of the 
data-line in a plot of y AH T /c versus w/c. The constant 



(w/c) is the value of w/c where the line crosses the 
w/c-axis. Both constants are determined empirically by 
drawing the line that best represents the experimental 
data. 



-86 



Having determined these constants for a given ce- 
ment , Equations 2 and 7 can be used to estimate the 
values of Q and AH' at paste concentrations not tested 
experimentally. Of course, for concentrations intermedi- 
ate between ones tested experimentally, rather good pre- 
dictions can be made directly from plots of Q versus w, 
and AH 1 versus w. Semi -logarithmic paper can be used 
to obtain approximately linear curves by plotting the Q, 
and AH 1 values on the log scales. 

Cement Density 

In Equation 2, dc is the density of cement as de- 
termined in water . This density is used directly in 
the equation and also in the calculation of c, though 
not in calculating O w which is the product of the sur- 
face area per gram (by Wagner turb.) and the cement 
density as determined in kerosene in the usual way. The 
use of two different determinations of cement density is 
a refinement that is not strictly necessary in practical 
work. Equation 2 already involves other approximations 
(see section 2); and the use throughout of the density 
as determined in kerosene — or, indeed, just an average 
cement density — would not make a large difference in the 
results. However, a relatively simple determination of 
the apparent density of cement in water can be made as 
.follows: 



Use proportions of cement and water appropriate for 
a bleeding test and such as do not cause appreciable 
entrainment of air. Mix as for a bleeding test, taking 
care that the final temperature shall be approximately 
room temperature. Pour the paste into an Erlenmeyer 
flask of nominal 300 ml capacity and obtain level meas- 
ure.* 



*A smali7~f lat glass plate can be used to strike off ex- 
cess paste and can also be used as a cover. It can be 
made close-fitting by grinding it against the rim of 
the flask with a little moistened abrasive powder 



prior to use. 



87- 



The total volume of the flask should be known accurately, 
and this volume, in cc, should he divided by the weight 
of the paste in the flask in grams to obtain the density 
of the paste, dm. From the quantities used in the mix 
calculate the ratio, r, of the weight of cement in grams 



volume of water in cc. Then 



d 

in 



dm 



= wd f + rw, or w = y - g 



rw m 

d ■ 



c 1 - w r + d „ - d m 

f m 



The apparent density of cement as deteimined in wa- 
ter is a little greater than the density as determined in 
kerosene because interaction between the cement and the 
water decreases the total volume of the system* (See 
section 7») 



-88- 




- 



~~ 



» *w